Upload Stable-diffusion using SD-Hub extension

.gitattributes

@@ -404,3 +404,8 @@ Stable-diffusion/projectIL_v3Vae.preview.png filter=lfs diff=lfs merge=lfs -text
 Stable-diffusion/projectIL_v4VAE.preview.png filter=lfs diff=lfs merge=lfs -text
 Stable-diffusion/projectIL_v5VAE.preview.png filter=lfs diff=lfs merge=lfs -text
 Stable-diffusion/rouwei_v080Epsilon.preview.png filter=lfs diff=lfs merge=lfs -text
+Stable-diffusion/illumiyumeXL_v35VPred.preview.png filter=lfs diff=lfs merge=lfs -text
+Stable-diffusion/rouwei_061Vpred.preview.png filter=lfs diff=lfs merge=lfs -text
+Stable-diffusion/rouwei_06Epsilon.preview.png filter=lfs diff=lfs merge=lfs -text
+Stable-diffusion/rouwei_v080Vpred.preview.png filter=lfs diff=lfs merge=lfs -text
+Stable-diffusion/rouwei_v080Vpred_trainingData.preview.png filter=lfs diff=lfs merge=lfs -text

Stable-diffusion/illumiyumeXL_v35VPred.json

@@ -0,0 +1,8 @@
+{
+    "sha256": "785B4B6DD828A0DA605543BB673FF9D107D5A68A7E452372EC3DC82595481DCB",
+    "modelId": 1308285,
+    "modelVersionId": 1887507,
+    "activation text": "",
+    "description": "IntroductionFor version 1.0:This model is based on 'Illustrious XL 1.0' with some minor modifications and was trained on the Danbooru2023 https://huggingface.co/datasets/nyanko7/danbooru2023 along with the dataset I previously used for training my LoRA models.For version 2.0:This developed model is intended to allow everyone to experience the v-pred version of Illustrious XL, instead of having to spend a large amount of STARDUST to unlock the Illustrious XL v3.0 v-pred and v3.5 v-pred versions.I independently researched and developed this version based on various existing XL model architectures. However, due to the many modifications I made, I\u2019m not sure it can still be considered 'Illustrious XL'.The model was trained on the danbooru2024 https://huggingface.co/datasets/deepghs/danbooru2024, danbooru_newest-all datasets, https://huggingface.co/datasets/deepghs/danbooru_newest-all as well as a custom dataset (which I collected and labeled using natural language with GPT-4.5, and later manually verified by me).I put a lot of time and effort into developing this version, so if you don't mind, please consider bidding on it so that others can use it through the CivitAI generator. Thank you all very much!For version 3.0:With this version, the model was created with the purpose of adapting to as many styles as possible, while also balancing detail stability in the generated images. This model includes styles and artist styles (from Danbooru and e621).Although it is oriented towards being a pre-trained model, you can use it normally. However, to achieve optimization, I suggest you combine it with LoRA or fine-tune it to create the style you desire.The model was trained on the danbooru2024 https://huggingface.co/datasets/deepghs/danbooru2024, danbooru_newest-all datasets https://huggingface.co/datasets/deepghs/danbooru_newest-all, e621 https://huggingface.co/datasets/deepghs/e621_newest?not-for-all-audiences=true as well as a custom dataset, with 40% of this data annotated using both tags and natural language.This model is an epsilon-prediction model that can easy to use.For version 3.1:This version improves the issues encountered in version 3.0. In addition, it also enhances image quality related to styles and artist styles (from Danbooru and e621).This model was trained on the same dataset as version 3.0, but I re-annotated it, added many new anime characters, and improved the quality of existing ones.The model improves stability when generating images at a resolution of 1536x1536.This version will have two variants: one for v-pred and one for e-pred (the e-pred version will be released first).For version 3.2:This model is a refined version of 3.1, incorporating hotfixes and enhancements. It features improved detailing in the eyes and more accurate anatomical proportions for the character.Additionally, the model demonstrates enhanced creativity and a better ability to accurately understand promptsThis model is also capable of generating images at large resolutions, e.g., 1024x2048 (I tested it and found the image quality to be quite decent). (Note: during training, I only trained it with images at a resolution of 1536x1536).For version 3.5:This model was trained on the Danbooru dataset, updated as of May 9th, 2025, with image sizes of 1536x1536.It fixes an important bug that appeared in version 2.0 of the v-pred variant.The model also improves stable style, anatomy, and prompt understanding compared to the previous version.Important NoteThis is the first base model I've created, so any feedback is welcome. Feel free to share your thoughts so I can improve it in future versions.Version 2.0 is a V-prediction model (unlike epsilon-prediction), and it requires a number of specific parameters.Version 3.0 should be set with a low CFG value, around 2 to 4. When you encounter images generated with high contrast (I don't know why CFG affect this, i will investigate and find the solution :v)Currently, the model is not available for use via Civitai Generation. You can visit the following website to use it:tensor.art https://tensor.art/models/835706876936308446Suggested settings:All example images were generated using the following settings:Positive prompt: masterpiece,best quality,amazing qualityNegative prompt: bad quality,worst quality,worst detail,sketch,censor, simple background,transparent backgroundCFG: 5-7 (For version 3.0 i suggest you should set this lower from 2-4 )Clip skip: 2Step: 20-30Sampler: Euler a/DPM++ 2S aNote: I don't use any post-processing and Lora to enhance the example images. I only use these settings and a custom prompt with my base model to generate.AcknowledgmentsThanks to narugo1992 https://huggingface.co/narugo and Nyanko https://huggingface.co/nyanko7 for sharing such valuable data.If you'd like to support my work, you can do so through Ko-fi https://ko-fi.com/duongve!",
+    "sd version": "Other"
+}

Stable-diffusion/illumiyumeXL_v35VPred.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:785b4b6dd828a0da605543bb673ff9d107d5a68a7e452372ec3dc82595481dcb
+size 6938040794

Stable-diffusion/rouwei_061Vpred.html

@@ -0,0 +1,250 @@
+
+                <head>
+                    
+                <meta charset="UTF-8">
+                    <link rel="stylesheet" type="text/css" href="/content/A1111/extensions/sd-civitai-browser-plus/style_html.css">
+                </head>
+                <div class="model-block">
+                    <h2><a href=https://civitai.com/models/950531 target="_blank" id="model_header">RouWei</a></h2>
+                    <h3 class="model-uploader">Uploaded by <a href="https://civitai.com/user/Minthybasis" target="_blank">Minthybasis</a><div class="avatar"><img src=https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/c3116f85-2b3b-4daf-a138-aed342f0cf2d/width=96/Minthybasis.jpeg></div></h3>
+                    <div class="civitai-version-info" style="display:flex; flex-wrap:wrap; justify-content:space-between;">
+                        <dl id="info_block">
+                            <dt>Version</dt>
+                            <dd>0.6 vpred</dd>
+                            <dt>Base Model</dt>
+                            <dd>Illustrious</dd>
+                            <dt>Published</dt>
+                            <dd>2024-11-15</dd>
+                            <dt>Availability</dt>
+                            <dd>Public</dd>
+                            <dt>CivitAI Tags</dt>
+                            <dd>
+                                <div class="civitai-tags-container">
+                                    <span class="civitai-tag">anime</span><span class="civitai-tag">base model</span>
+                                </div>
+                            </dd>
+                            <dt>Download Link</dt>
+                            <dd><a href=https://civitai.com/api/download/models/1064242 target="_blank">https://civitai.com/api/download/models/1064242</a></dd>
+                        </dl>
+                        <div style="align-self:center; min-width:320px;">
+                            <div>
+                                <p style="line-height: 2; font-weight: bold;"><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Use the model without crediting the creator<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Sell images they generate<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Run on services that generate images for money<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Run on Civitai<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Share merges using this model<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="red"><path d="M18 6l-12 12"></path><path d="M6 6l12 12"></path></svg></span> Sell this model or merges using this model<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="red"><path d="M18 6l-12 12"></path><path d="M6 6l12 12"></path></svg></span> Have different permissions when sharing merges</p>
+                            </div>
+                        </div>
+                    </div>
+                    <input type="checkbox" id="preview-civitai-description" class="description-toggle-checkbox">
+                    <div class="model-description">
+                        <h2>Description</h2>
+                        <h2 id="in-depth-retraining-of-illustrious-to-achieve-best-prompt-adherence-knowledge-and-state-of-the-art-performance.-66mudratg">In depth retraining of Illustrious to achieve best prompt adherence, knowledge and state of the art performance.</h2><p><span style="color:rgb(134, 142, 150)">Big dreams come true</span></p><p><span style="color:rgb(134, 142, 150)">The version number is just an index of current final release, not a fraction of the planned training.</span></p><p><a target="_blank" rel="ugc" href="https://huggingface.co/Minthy/RouWei-0.8"><span style="color:rgb(134, 142, 150)">HF repo</span></a></p><p></p><h3 id="large-scale-finetune-using-gpu-cluster-with-a-dataset-of-~13m-pictures-(~4m-with-natural-text-captions)-pqrwpp4f0">Large scale finetune using gpu cluster with a dataset of ~13M pictures (~4M with natural text captions)</h3><ul><li><p>Fresh and wast knowledge about characters, concepts, styles, cultural and related things</p></li><li><p>The best prompt adherence among SDXL anime models at the moment of release</p></li><li><p>Solved main problems with tags bleeding and biases, common for Illustrious, NoobAi and other checkpoints</p></li><li><p>Excellent aesthetics and knowledge across a wide range of styles (<a target="_blank" rel="ugc" href="https://files.catbox.moe/anleso.txt">over 50,000 artists</a> (<a rel="ugc" href="https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg">examples</a>), including hundreds of unique cherry-picked datasets from private galleries, including those received from the artists themselves)</p></li><li><p>High flexibility and variety without stability tradeoff</p></li><li><p>No more annoying watermarks for popular styles thanks to clean dataset</p></li><li><p>Vibrant colors and smooth gradients without trace of burning, full range even with epsilon</p></li><li><p>Pure training from Illustrious v0.1 without involving third-party checkpoints, Loras, tweakers, etc.</p></li></ul><p></p><h3 id="there-are-also-some-issues-and-changes-compared-to-the-previous-version-please-rtfm.-aaa28dbe6">There are also some issues and changes compared to the previous version, please RTFM.</h3><p></p><p>Dataset cut-off - end of April 2025.</p><p></p><h1 id="features-and-prompting:-j89sanl8e">Features and prompting:</h1><p></p><h2 id="important-change:-f6w18r0i0">Important change:</h2><p>When you are prompting artist styles, especially mixing several, <strong><u>their tags MUST BE in a separate CLIP chunk</u></strong>. Just add BREAK after it (for A1111 and derivatives), use conditioning concat node (for Comfy) <strong><u>or at least put them in the very end</u></strong>. Otherwise, significant degradation of results is likely.</p><p></p><h2 id="basic:-aroav4616">Basic:</h2><p></p><p>The checkpoint works both with short-simple and long-complex prompts. However, if there are contradictory or weird things - unlike with others they won't be ignored affecting the output. No guide-rails, no safeguards, no lobotomy.</p><p>Just prompt what you want to see and don't prompt what shouldn't be on the picture. If you want to have a view from above - don't put ceiling into positive, if you want to have crop view with head out of frame - don't make detailed description of character facial features, and so on. Pretty simple but sometimes people are missing it.</p><p></p><p>Version 0.8 comes with advanced understanding of natural text prompts. <u>It doesn't mean that you are obligated to use it</u>, tags only - completely fine, especially because understanding of tags combinations is also improved.</p><p>Do not expect it to perform like Flux or other models based on T5 or LLM text encoders. The whole size ot SDXL checkpoint is less then only that text encoder, in addition illustrious-v0.1 which is used as the base completely forgot a lot of general things from vanilla sdxl-base.</p><p>However, even in current state it works much better, allows to do new things usually impossible without external guidance, as well making manual editing, inpainting, etc more convenient.</p><p></p><p>To achieve best performance you should keep track of CLIP chunks. In SDXL the prompt is separated into a chunks of 75 (77 including BOS and EOS) tokens, that are processing by CLIP separately, and only then are concatinating and comes as conditions to unet.</p><p>If you want to specify some features for character/object and separate them from other prompt parts - make sure they are in the same chunk and optionally separate it with <code style="text-wrap: wrap">BREAK</code>. It will not solve problem of traits mixing completely, but can reduce it improving overall understanding, since text encoders on RouWei are able to process the whole sequence, not individual concepts better then others.</p><p>Dataset contains only booru-style tags and natural text expressions. Despite having a share of furries, real life photos, western media, etc. all captions have been converted to classic booru style to avoid a number of problems from mixing of different systems. So e621 tags won't be understanded properly.</p><p></p><h3 id="sampling-parameters:-7andwk13p">Sampling parameters:</h3><ul><li><p>~1 megapixel for txt2img, any AR with resolution multiple of 32 (1024x1024, 1056x, 1152x, 1216x832,...). Euler_a, 20..28steps.</p></li><li><p>CFG: for epsilon version 4..9 (7 is best), for vpred version, 3..5</p></li><li><p>Sigmas multiply may improve results a bit, CFG++ samplers work fine. LCM/PCM/DMD/... and exotic samplers untested.</p></li><li><p><strong>Some schedulers doesn't work well.</strong></p></li><li><p>Highresfix - x1.5 latent + denoise 0.6 or any gan + denoise 0.3..0.55.</p></li><li><p>For vpred version lower CFG 3..5 is needed!</p></li></ul><p></p><h3 id="for-vpred-version-lower-cfg-3..5-is-needed!-uwz9s21js"><strong>For vpred version lower CFG 3..5 is needed!</strong></h3><p></p><h2 id="quality-classification:-kw87tas72">Quality classification:</h2><p>Only 4 quality tags:</p><pre><code style="text-wrap: wrap">masterpiece, best quality</code></pre><p>for positive and</p><pre><code style="text-wrap: wrap">low quality, worst quality</code></pre><p>for negative.</p><p>Nothing else. Actually you can even omit positive and reduce negative to <code style="text-wrap: wrap">low quality</code> only, since they can affect basic style and composition.</p><p>Meta tags like <code style="text-wrap: wrap">lowres</code> have been removed and don't work, better not to use them. Low resolution images have been either removed or upscaled and cleaned with DAT depending on their importance.</p><p></p><h2 id="negative-prompt:-79eot1hhx">Negative prompt:</h2><pre><code style="text-wrap: wrap">worst quality, low quality, watermark</code></pre><p>That's all, no need of "rusty trombone", "farting on prey" and others. Do not put tags like greyscale, monochrome in negative unless you understand what are you doing. Extra tags for brightness/colors/contrast section below can be used</p><p></p><h2 id="artist-styles:-6yumq5hap">Artist styles:</h2><p></p><p><a target="_blank" rel="ugc" href="https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg">Grids with examples</a>, <a target="_blank" rel="ugc" href="https://files.catbox.moe/anleso.txt">list/wildcard</a> (also can be found in "training data").</p><p></p><h3 id="used-with-&quot;by-&quot;-it's-mandatory.-it-will-not-work-properly-without-it.-innc8cnap"><u>Used with "by " it's mandatory</u>. It will not work properly without it.</h3><p>"by " is a meta-token for styles to avoid mixing/misinterpret with tags/characters of similar or close name. This allows to have a better results for styles and at the same time avoid random style fluctuation that you may observe in other checkpoints.</p><p></p><p>Multiple give very interesting results, can be controlled with prompt weights and spells.</p><p></p><h3 id="you-must-add-break-after-artistsstyle-tags-(for-a1111)-or-concat-conditioning-(for-comfy)-or-put-them-in-the-very-end-of-your-prompt.-36i67w9c0">YOU MUST ADD <code style="text-wrap: wrap">BREAK</code> after artists/style tags (for A1111) or concat conditioning (for Comfy) or put them in the very end of your prompt.</h3><p>For example:</p><pre><code style="text-wrap: wrap">by kantoku, by wlop, best quality, masterpiece BREAK 1girl, ...</code></pre><p></p><h2 id="general-styles:-5ig9sgqza">General styles:</h2><pre><code style="text-wrap: wrap">2.5d, anime screencap, bold line, sketch, cgi, digital painting, flat colors, smooth shading, minimalistic, ink style, oil style, pastel style</code></pre><h3 id="booru-tags-styles:-uq05a7u5y">Booru tags styles:</h3><pre><code style="text-wrap: wrap">1950s (style), 1960s (style), 1970s (style), 1980s (style), 1990s (style), 2000s (style), animification, art nouveau, pinup (style), toon (style), western comics (style), nihonga, shikishi, minimalism, fine art parody</code></pre><p>and everything from <a target="_blank" rel="ugc" href="https://danbooru.donmai.us/wiki_pages/traditional_media">this group</a>.</p><p>Can be used in combinations (with artists too), with weights, both in positive and negative prompts.</p><p></p><h2 id="characters:-8htyrng9m">Characters:</h2><p>Use full name booru tag and proper formatting, like <code style="text-wrap: wrap">karin_(blue_archive)</code> -&gt; <code style="text-wrap: wrap">karin \(blue archive\)</code>, use skin tags for better reproducing, like <code style="text-wrap: wrap">karin \(bunny\) \(blue archive\)</code>. Autocomplete extension might be very useful.</p><p>Most characters are recognized just by their booru tag, but it will be more accurate if you describe their basic traits. Here you can easily redress your waifu/husbendo just by the prompt without suffering from the typical leaks of basic features.</p><p></p><h2 id="natural-text:-cq8whprlj">Natural text:</h2><p>Use it in combination with booru tags, works great. Use only natural text after typing styles and quality tags. Use just booru tags and forget about it, it's all up to you. To get best performance keep track if CLIP 75 tokens chunks.</p><p></p><p>About 4M of images in dataset had hybrid natural-text captions, made by Claude, GPT, Gemini, <a target="_blank" rel="ugc" href="https://huggingface.co/Minthy/ToriiGate-v0.4-7B">ToriiGate</a>, then refactored, cleaned and combined with tags in different variations for augmentation.</p><p></p><p>Unlike typical captions, these contains character names which is very useful. Better to keep it clean, short and convenient description works best. Better <strong><u>not</u></strong> use long and sloppy BS like</p><pre><code style="text-wrap: wrap">A mysteriously enchanting feminine entity of indeterminate yet youthful essence, whose celestial visage radiates with the ethereal luminescence of a thousand dying stars, blessed with locks cascading like the golden rivers of ancient mythology, perhaps styled in a manner reminiscent of contemporary fashion trends though not necessarily adhering to any specific aesthetic paradigm. Her eyes, pools of unfathomable depth and hue, sparkle with the wisdom of millennia yet maintain an innocent quality that defies temporal constraints...</code></pre><p>For captioning you can use <a target="_blank" rel="ugc" href="https://huggingface.co/Minthy/ToriiGate-v0.4-7B">ToriiGate</a> in short mode.</p><p></p><p>And don't expect it to be as good as flux and others, it tries very hard and after several rolls usually you can get what you want, but it is not that stable and detailed.</p><p></p><h2 id="lots-of-tailears-related-concepts:-qsxmk9gzm">Lots of Tail/Ears-related concepts:</h2><p>Oh yeah</p><pre><code style="text-wrap: wrap">tail censor, holding own tail, hugging own tail, holding another's tail, tail grab, tail raised, tail down, ears down, hand on own ear, tail around own leg, tail around penis, tailjob, tail through clothes, tail under clothes, lifted by tail, tail biting, tail penetration (including a specific indication of vaginal/anal), tail masturbation, holding with tail, panties on tail, bra on tail, tail focus, presenting own tail...</code></pre><p>(booru meaning, not e621) and many others with natural text. The majority works perfectly, some requires a lot of rolling.</p><p></p><h1 id="brightnesscolorscontrast:-lk2g7nwtj">Brightness/colors/contrast:</h1><p>You can use extra meta tags to control it:</p><pre><code style="text-wrap: wrap">low brightness, high brightness, low saturation, high saturation, low gamma, high gamma, sharp colors, soft colors, hdr, sdr</code></pre><p><a target="_blank" rel="ugc" href="https://files.catbox.moe/r8wrb6.jpg">Example</a></p><p>They work both in epsilon and vpred version and works really good.</p><p>Epsilon version relies on them too much. Without low brightness or low gamma or limited range (in negative) it might be difficult to achieve true 0,0,0 black, the same often true for white.</p><p>Both epsilon and vpred versions have like true zsnr, full range of colors and brightness without common flaws observed. But they behaves differently, just try it.</p><p></p><h1 id="vpred-version-9pr2w5h4d">Vpred version</h1><p></p><p>Main thing you need to know - lower your CFG from 7 down to 5 (or less). Otherwise, the use is similar with advantages.</p><p>It seems that starting from v0.7 vpred works flawlessly now. It shouldn't suffer from ignorance of tags close to the 75tokens chunk borders like nai. It is more difficult to get burned images - even on cfg7 usually it just over-saturated but with smooth gradients, which can be useful for some styles. Yes it can make anything from (0,0,0) to (255,255,255). You will find brightness meta tags described above quite useful for easier/lazy prompting, natural text expressions also work. To get the most dark image - put <code style="text-wrap: wrap">high brightness</code> into <u>negative</u> and/or use <code style="text-wrap: wrap">low brightness, low gamma</code> tags. If you don't like very bright skin on dark background and want to reduce contrast (or on the contrary, enhance the effect) - use hdr/sdr in negative/positive.</p><p>It was reported that in rare cases on some prompts there is a drop in contrast. Looks like other vpred models have same behaviour with such prompts, adding a "separator" closer to the border of the 75-token chunk fixes this. However, with 0.7 I haven't encountered this myself.</p><p></p><p><span style="color:rgb(134, 142, 150)">To launch vpred version you will need dev build of A1111, Comfy (with special loader node), Forge or Reforge. Just use same parameters (Euler a, cfg 3..5, 20..28 steps) like epsilon. No need to use Cfg rescale, but you can try it, cfg++ works great.</span></p><p></p><h1 id="base-model:-lwa0dfdbf">Base model:</h1><p>The model here has a small unet polishint after main training to improve small details, bump up resolution and others. Hovewer, you may be also interested into a RouWei-Base, which sometimes can perform better at complex prompts despite having minor mistakes in small details. It also comes in FP32, for example if you want to use fp32 text encoder nodes in Comfy, merge it or finetune.</p><p>It can be found in <a target="_blank" rel="ugc" href="https://huggingface.co/Minthy/RouWei-0.8">Huggingface repo</a></p><p></p><h1 id="known-issues:-jxgt6yemu">Known issues:</h1><p>Off course there are:</p><ul><li><p>Artists and style tags must be seperated into a different chunk from main prompt or come very last</p></li><li><p>There may be some positional or combinational bias in rare cases, but it's not yet clear.</p></li><li><p>There are some complaints about few of the general styles.</p></li><li><p>Epsilon version relies too much on brightness meta tags, sometimes you will need to use them to get desired brightness shift</p></li><li><p>Some newly added styles/characters might be not as good and disctinct as they deserve to</p></li><li><p>To be discovered</p><p></p></li><li><p></p></li></ul><h3 id="requests-for-artistscharacters-in-future-models-are-open.-if-you-find-artistcharacterconcept-that-perform-weak-inaccurate-or-has-strong-watermark-please-report-will-add-them-explicitly.-follow-for-a-new-versions.-kgc39ed4e">Requests for artists/characters in future models are open. If you find artist/character/concept that perform weak, inaccurate or has strong watermark - please report, will add them explicitly. Follow for a new versions.</h3><p></p><h2 id="join-the-discord-server-9lpaojolv"><a target="_blank" rel="ugc" href="https://discord.gg/ZXHENAhqE9">JOIN THE DISCORD SERVER</a></h2><p></p><h2 id="license:-0hpy342uh">License:</h2><p>Same as illustrious. Fell free to use in your merges, finetunes, ets. but please leave a link or mention, it is mandatory</p><p></p><h2 id="how-it's-made-zc68i8qve">How it's made</h2><p>I'll consider to make a report or something like it later. For sure.</p><p>In short, 98% of work is related to dataset preparations. Instead of blindly relying on loss-weighting based on tag frequency from nai paper, a custom guided loss-weighting implementation along with asynchronous collator for balancing have been used. Ztsnr (or close to it) with Epsilon prediction was achieved using noise scheduler augmentation.</p><p></p><p>Spent compute - over 8k hours of H100 (apart from research and fail attempts)</p><p></p><h1 id="thanks:-4tsy2pj9q">Thanks:</h1><p>First of all I'd like to acknowledge everyone who supports open source, develops in improves code. Thanks to the authors of illustrious for releasing model, thank to NoobAI team for being pioneers in open finetuning of such a scale, sharing experience, raising and solving issues that previously went unnoticed.</p><h3 id="personal:-9a91vqx2i">Personal:</h3><p>Artists wish to remain anonymous for sharing private works; Few anonymous persons - donations, code, captions, etc., Soviet Cat - GPU sponsoring; Sv1. - llm access, captioning, code; K. - training code; Bakariso - datasets, testing, advices, insides; NeuroSenko - donations, testing, code; LOL2024 - a lot of unique datasets; T.,[] - datasets, testing, advises; rred, dga, Fi., ello - donations; TekeshiX - datasets. And other fellow brothers that helped. Love you so much ❤️.</p><p>And off course everyone who made feedback and requests, it's really valuable.</p><p><span style="color:rgb(134, 142, 150)">If I forgot to mention anyone, please notify.</span></p><p></p><h2 id="donations-jdo27wd0z">Donations</h2><p>If you want to support - share my models, leave feedback, make a cute picture with kemonomimi-girl. And of course, support original artists.</p><p>AI is my hobby, I'm spending money on it and not begging for donations. However, it has turned into a large-scale and expensive undertaking. Consider to support to accelerate new training and researches.</p><p><span style="color:rgb(134, 142, 150)">(Just keep in mind that I can waste it on alcohol or cosplay girls)</span></p><p>BTC: bc1qwv83ggq8rvv07uk6dv4njs0j3yygj3aax4wg6c</p><p>ETH/USDT(e): 0x04C8a749F49aE8a56CB84cF0C99CD9E92eDB17db</p><p>XMR: 47F7JAyKP8tMBtzwxpoZsUVB8wzg2VrbtDKBice9FAS1FikbHEXXPof4PAb42CQ5ch8p8Hs4RvJuzPHDtaVSdQzD6ZbA5TZ</p><p>if you can offer gpu-time (a100+) - PM.</p><p></p><p></p><p></p><p></p><p></p>
+                    </div>
+                    <label for="preview-civitai-description" class="description-toggle-label"></label>
+                </div>
+                <div align=center><div class="sampleimgs"><input type="radio" name="zoomRadio" id="resetZoom" class="zoom-radio" checked>
+                    <div class="model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_0" class="zoom-radio">
+                    <label for="zoomRadiopreview_0" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/f96c7ce9-0eaf-4461-bee3-176a008ee8d2/width=1536/40258578.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/f96c7ce9-0eaf-4461-bee3-176a008ee8d2/width=1536/40258578.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>masterpiece, best quality, good quality,
+1girl, konpaku youmu, half-closed eyes, solo, black hairband, (konpaku youmu (ghost):1.2), hairband, green vest, puffy sleeves, upper body, bow, black bowtie, black bow, short hair, vest, bowtie, ribbon, leaf, shirt, wind, short sleeves, black background, looking down, white shirt, hitodama, black ribbon, hair ribbon, puffy short sleeves, grey hair, simple background, floating hair, closed mouth, expressionless, wing collar, hitodama print, buttons, green eyes, collared shirt, falling leaves, aqua eyes, glowing eyes, pale skin, (skeleton, split screen, straight-on, android, crank, gauge, gears, glowing, light bulb, mechabare, sparkle, x-ray:1.2), 
+&lt;lora:youmuanon_noob-vpred0.1_flawless_ds_v2:1&gt;</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, bad quality, lowres, jpeg artifacts, bad anatomy, bad hands, multiple views, signature, watermark, censored, covered navel, limited range, high brightness, high contrast,</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>4147779862</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1024x1024</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred_rc1</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>5</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_0">
+                                    <label class="tab-label" for="chckpreview_0">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl-vae-fp16-fix.safetensors</dd></div><div class="civitai-meta"><dt>Ensd</dt><dd>31337</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;62c7c729ad&#x27;, &#x27;model&#x27;: &#x27;e63d568c63&#x27;, &#x27;lora:youmuanon_noob-vpred0.1_flawless_ds_v2&#x27;: &#x27;7b7e860a9bae&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.3.6dev2-v1.10.1RC-latest-1453-g0fd79513</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;7b7e860a9bae&#x27;, &#x27;name&#x27;: &#x27;youmuanon_noob-vpred0.1_flawless_ds_v2&#x27;, &#x27;type&#x27;: &#x27;lora&#x27;, &#x27;weight&#x27;: 1}, {&#x27;hash&#x27;: &#x27;e63d568c63&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred_rc1&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>e63d568c63</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Align Your Steps</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>2x-AnimeSharpV3</dd></div><div class="civitai-meta"><dt>Adetailer model</dt><dd>face_yolov8n.pt</dd></div><div class="civitai-meta"><dt>Adetailer version</dt><dd>24.9.0</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Sigmas_merge_mode</dt><dd>Multiply</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.5</dd></div><div class="civitai-meta"><dt>Sigmas_mult_factor</dt><dd>2</dd></div><div class="civitai-meta"><dt>Adetailer mask blur</dt><dd>8</dd></div><div class="civitai-meta"><dt>Adetailer confidence</dt><dd>0.3</dd></div><div class="civitai-meta"><dt>Sigmas_merge_enabled</dt><dd>True</dd></div><div class="civitai-meta"><dt>Adetailer dilate erode</dt><dd>4</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Adetailer inpaint padding</dt><dd>32</dd></div><div class="civitai-meta"><dt>Adetailer negative prompt</dt><dd>{}</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div><div class="civitai-meta"><dt>Adetailer denoising strength</dt><dd>0.4</dd></div><div class="civitai-meta"><dt>Old prompt editing timelines</dt><dd>True</dd></div><div class="civitai-meta"><dt>Adetailer inpaint only masked</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_1" class="zoom-radio">
+                    <label for="zoomRadiopreview_1" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/76198546-76cd-4bf4-9fc8-f935da633bbf/width=1328/40258585.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/76198546-76cd-4bf4-9fc8-f935da633bbf/width=1328/40258585.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>(by deel \(rkeg\):0.8), best quality, masterpiece,
+1girl, folinic \(arknights\), cat girl, brown hair, multicolored hair, cat ears, surprised, blush, hair flower, holding papers, BREAK
+indoors, scientific lab, glass floor, intricate interior, computers, research facility</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, watermark</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>2290774044</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>832x1216</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_1">
+                                    <label class="tab-label" for="chckpreview_1">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;84c8888446&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;84c8888446&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>84c8888446</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.6</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_2" class="zoom-radio">
+                    <label for="zoomRadiopreview_2" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/5033de43-81ec-418e-9429-f2a7425353c4/width=1728/40258592.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/5033de43-81ec-418e-9429-f2a7425353c4/width=1728/40258592.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by ratatatat74, by tarakanovich,cube85, masterpiece, best quality,
+1girl, fake animal ears, cat ears, long black hair, large breasts, erect nipples, completely nude, butt plug, black thighhighs, garter straps, squatting, reverse cowgirl position, girl on top, from behind, ass, mature female, pussy juice, ass grab, large penis, vaginal, pov hands, tally, (slap mark:0.7),
+indoors, bedroom, dark night, dimly lit room, dark blue theme, low brightness, low gamma, sdr</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, limited range</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>1178610355</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1152x896</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_2">
+                                    <label class="tab-label" for="chckpreview_2">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_3" class="zoom-radio">
+                    <label for="zoomRadiopreview_3" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/9af10cd7-4513-4147-a4f2-033f46afd3e0/width=1824/40258577.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/9af10cd7-4513-4147-a4f2-033f46afd3e0/width=1824/40258577.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>(by ratatatat74:0.7), by tarakanovich, cube85, masterpiece, best quality,
+1girl, vulpisfoglia \(arknights\), fox girl, fox ears, large ears, ears down, blonde hair, large breasts, yellow eyes, fox tail, nightgown, see-through, bottomless, from behind, tail censor, (scar on animal ear:0.8), white panties, naughty face,
+all fours, fox shadow puppet, glowing eyes,
+black background, low brightness, low gamma</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, extra ears</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>1556719420</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1216x832</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>20</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_3">
+                                    <label class="tab-label" for="chckpreview_3">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_4" class="zoom-radio">
+                    <label for="zoomRadiopreview_4" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/0d48a8dd-f3d3-42f6-b79d-de7de7b52139/width=1200/40258590.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/0d48a8dd-f3d3-42f6-b79d-de7de7b52139/width=1200/40258590.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>masterpiece, best quality, by merrytail, by sub-res, by kie \(wylee2212\),
+1girl, fox devil \(chainsaw man\), fox girl, fox ears, fox tail, whisker markings, long white hair, red eyes, (extra eyes:1.3), fangs, evil smile, white kimono with floral print, collarbone, hair bun, fox shadow puppet, cowboy shot, posing,
+black background, low brightness</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, limited range, monster, 1other, animal, fox</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>1376750140</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>800x1344</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_4">
+                                    <label class="tab-label" for="chckpreview_4">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_5" class="zoom-radio">
+                    <label for="zoomRadiopreview_5" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/209ff688-e01d-41d4-b842-a3ffadd3741f/width=1824/40258575.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/209ff688-e01d-41d4-b842-a3ffadd3741f/width=1824/40258575.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>masterpiece, best quality, by done \(donezumi\),
+1girl, kon \(tokyo ravens\), tokyo ravens, fox girl, fox ears, fox tail, white hair, thick eyebrows, pale skin, blue eyes, white aqua japanese clothes, detached sleeves, ribbon trim, fox shadow puppet, pom pom \(clothes\), black shorts, white thighhighs, hips, happy, BREAK floating in air, legs apart, posing, full body, wide shot,
+white background, high brightness</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>2168859150</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1216x832</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>20</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_5">
+                                    <label class="tab-label" for="chckpreview_5">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;84c8888446&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;84c8888446&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>84c8888446</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_6" class="zoom-radio">
+                    <label for="zoomRadiopreview_6" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/90435b3a-ef7d-4be2-b700-00601088c553/width=1248/40258594.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/90435b3a-ef7d-4be2-b700-00601088c553/width=1248/40258594.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by tandohark, (by personal ami:0.6), best quality, masterpiece,
+2girls, asymmetrical docking BREAK
+izuna \(blue archive\), fox girl, fox ears, fox mask, fox tail, halo, topless, denim shorts, striped bikini bra, medium breasts, sideboob, visors cap BREAK
+wakamo \(blue archive\), fox girl, fox ears, fox tail, long black hair, red halo, hair ornament, large breasts, fox mask, white bikini swimsuit, collarbone, :3 BREAK
+outdoors, beach, seashore, palms, clear blue sky</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, watermark</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>1114850771</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>832x1216</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>5</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_6">
+                                    <label class="tab-label" for="chckpreview_6">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;84c8888446&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;84c8888446&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>84c8888446</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_7" class="zoom-radio">
+                    <label for="zoomRadiopreview_7" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/23c627f5-451b-47de-a1d3-7dce53adb8d9/width=1248/40258589.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/23c627f5-451b-47de-a1d3-7dce53adb8d9/width=1248/40258589.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by akai sashimi, by nat the lich, by foxykuro, by nekoya \(liu\), masterpiece,  best quality,
+1girl, kal&#x27;tsit \(arknights\), cat girl, lynx ears, silver hair, green eyes/, naked towel, covered nipples, off shoulder, oripathy lesion \(arknights\), collarbone, shaded face, (evil smile:1.1), smug, open mouth, angry, fang, tilted head, wet,
+holding handcuffs,
+cowboy shot, motion lines,
+indoors, office, dimmed, dark background</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, watermark, dutch angle</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>2542436241</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>832x1216</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_7">
+                                    <label class="tab-label" for="chckpreview_7">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Ensd</dt><dd>31337</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_8" class="zoom-radio">
+                    <label for="zoomRadiopreview_8" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/fd31b937-db26-404f-ba61-c86eecedb37c/width=1328/40258587.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/fd31b937-db26-404f-ba61-c86eecedb37c/width=1328/40258587.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by akai sashimi, by nat the lich, by foxykuro, by nekoya \(liu\), masterpiece, best quality,
+1girl, kal&#x27;tsit \(arknights\), cat girl, lynx ears, silver hair, green eyes, short tail, completely nude, oripathy lesion \(arknights\), scars on spine, reverse cowgirl position, motion lines, from below, open mouth, smug, heavenly ass, v,
+1boy, vaginal, sex, pov, deep penetration, pussy juice trail,
+indoors, bedroom, dimmed, dark background</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, watermark, abstract background</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>3956367149</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>832x1216</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>20</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_8">
+                                    <label class="tab-label" for="chckpreview_8">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.6</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_9" class="zoom-radio">
+                    <label for="zoomRadiopreview_9" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/11c49fc0-f6fa-40b2-bb86-30e3a2c9cd0a/width=2064/40258599.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/11c49fc0-f6fa-40b2-bb86-30e3a2c9cd0a/width=2064/40258599.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by kate fox, masterpiece, best quality,
+1girl, yuel \(granblue fantasy\), fox girl, yellow eyes, fox ears, large fox tail, bikini bra, fur trim, elbow gloves, :3, wide shot, BREAK
+outdoors, torii, asian nature, epic landscape of mountains, above clouds</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, bad hands, watermark
+worst quality, low quality</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>2135703813</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1376x720</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>20</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_9">
+                                    <label class="tab-label" for="chckpreview_9">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;84c8888446&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;84c8888446&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>84c8888446</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Rescale_cfg_enabled</dt><dd>True</dd></div><div class="civitai-meta"><dt>Rescale_cfg_multiplier</dt><dd>0.7</dd></div></div></div></div></dl></div></div></div></div>
+                

Stable-diffusion/rouwei_061Vpred.json

@@ -0,0 +1,8 @@
+{
+    "sha256": "0EC35D96FC8C6F1B205222077A7D5E929DD2E8D1DEBB2E130AD2E94E6CB21383",
+    "modelId": 950531,
+    "modelVersionId": 1130140,
+    "activation text": "",
+    "description": "In depth retraining of Illustrious to achieve best prompt adherence, knowledge and state of the art performance.Big dreams come trueThe version number is just an index of current final release, not a fraction of the planned training.HF repo https://huggingface.co/Minthy/RouWei-0.8Large scale finetune using gpu cluster with a dataset of ~13M pictures (~4M with natural text captions)Fresh and wast knowledge about characters, concepts, styles, cultural and related thingsThe best prompt adherence among SDXL anime models at the moment of releaseSolved main problems with tags bleeding and biases, common for Illustrious, NoobAi and other checkpointsExcellent aesthetics and knowledge across a wide range of styles (over 50,000 artists https://files.catbox.moe/anleso.txt (examples https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg), including hundreds of unique cherry-picked datasets from private galleries, including those received from the artists themselves)High flexibility and variety without stability tradeoffNo more annoying watermarks for popular styles thanks to clean datasetVibrant colors and smooth gradients without trace of burning, full range even with epsilonPure training from Illustrious v0.1 without involving third-party checkpoints, Loras, tweakers, etc.There are also some issues and changes compared to the previous version, please RTFM.Dataset cut-off - end of April 2025.Features and prompting:Important change:When you are prompting artist styles, especially mixing several, their tags MUST BE in a separate CLIP chunk. Just add BREAK after it (for A1111 and derivatives), use conditioning concat node (for Comfy) or at least put them in the very end. Otherwise, significant degradation of results is likely.Basic:The checkpoint works both with short-simple and long-complex prompts. However, if there are contradictory or weird things - unlike with others they won't be ignored affecting the output. No guide-rails, no safeguards, no lobotomy.Just prompt what you want to see and don't prompt what shouldn't be on the picture. If you want to have a view from above - don't put ceiling into positive, if you want to have crop view with head out of frame - don't make detailed description of character facial features, and so on. Pretty simple but sometimes people are missing it.Version 0.8 comes with advanced understanding of natural text prompts. It doesn't mean that you are obligated to use it, tags only - completely fine, especially because understanding of tags combinations is also improved.Do not expect it to perform like Flux or other models based on T5 or LLM text encoders. The whole size ot SDXL checkpoint is less then only that text encoder, in addition illustrious-v0.1 which is used as the base completely forgot a lot of general things from vanilla sdxl-base.However, even in current state it works much better, allows to do new things usually impossible without external guidance, as well making manual editing, inpainting, etc more convenient.To achieve best performance you should keep track of CLIP chunks. In SDXL the prompt is separated into a chunks of 75 (77 including BOS and EOS) tokens, that are processing by CLIP separately, and only then are concatinating and comes as conditions to unet.If you want to specify some features for character/object and separate them from other prompt parts - make sure they are in the same chunk and optionally separate it with BREAK. It will not solve problem of traits mixing completely, but can reduce it improving overall understanding, since text encoders on RouWei are able to process the whole sequence, not individual concepts better then others.Dataset contains only booru-style tags and natural text expressions. Despite having a share of furries, real life photos, western media, etc. all captions have been converted to classic booru style to avoid a number of problems from mixing of different systems. So e621 tags won't be understanded properly.Sampling parameters:~1 megapixel for txt2img, any AR with resolution multiple of 32 (1024x1024, 1056x, 1152x, 1216x832,...). Euler_a, 20..28steps.CFG: for epsilon version 4..9 (7 is best), for vpred version, 3..5Sigmas multiply may improve results a bit, CFG++ samplers work fine. LCM/PCM/DMD/... and exotic samplers untested.Some schedulers doesn't work well.Highresfix - x1.5 latent + denoise 0.6 or any gan + denoise 0.3..0.55.For vpred version lower CFG 3..5 is needed!For vpred version lower CFG 3..5 is needed!Quality classification:Only 4 quality tags:masterpiece, best qualityfor positive andlow quality, worst qualityfor negative.Nothing else. Actually you can even omit positive and reduce negative to low quality only, since they can affect basic style and composition.Meta tags like lowres have been removed and don't work, better not to use them. Low resolution images have been either removed or upscaled and cleaned with DAT depending on their importance.Negative prompt:worst quality, low quality, watermarkThat's all, no need of \"rusty trombone\", \"farting on prey\" and others. Do not put tags like greyscale, monochrome in negative unless you understand what are you doing. Extra tags for brightness/colors/contrast section below can be usedArtist styles:Grids with examples https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg, list/wildcard https://files.catbox.moe/anleso.txt (also can be found in \"training data\").Used with \"by \" it's mandatory. It will not work properly without it.\"by \" is a meta-token for styles to avoid mixing/misinterpret with tags/characters of similar or close name. This allows to have a better results for styles and at the same time avoid random style fluctuation that you may observe in other checkpoints.Multiple give very interesting results, can be controlled with prompt weights and spells.YOU MUST ADD BREAK after artists/style tags (for A1111) or concat conditioning (for Comfy) or put them in the very end of your prompt.For example:by kantoku, by wlop, best quality, masterpiece BREAK 1girl, ...General styles:2.5d, anime screencap, bold line, sketch, cgi, digital painting, flat colors, smooth shading, minimalistic, ink style, oil style, pastel styleBooru tags styles:1950s (style), 1960s (style), 1970s (style), 1980s (style), 1990s (style), 2000s (style), animification, art nouveau, pinup (style), toon (style), western comics (style), nihonga, shikishi, minimalism, fine art parodyand everything from this group https://danbooru.donmai.us/wiki_pages/traditional_media.Can be used in combinations (with artists too), with weights, both in positive and negative prompts.Characters:Use full name booru tag and proper formatting, like karin_(blue_archive) -> karin \\(blue archive\\), use skin tags for better reproducing, like karin \\(bunny\\) \\(blue archive\\). Autocomplete extension might be very useful.Most characters are recognized just by their booru tag, but it will be more accurate if you describe their basic traits. Here you can easily redress your waifu/husbendo just by the prompt without suffering from the typical leaks of basic features.Natural text:Use it in combination with booru tags, works great. Use only natural text after typing styles and quality tags. Use just booru tags and forget about it, it's all up to you. To get best performance keep track if CLIP 75 tokens chunks.About 4M of images in dataset had hybrid natural-text captions, made by Claude, GPT, Gemini, ToriiGate https://huggingface.co/Minthy/ToriiGate-v0.4-7B, then refactored, cleaned and combined with tags in different variations for augmentation.Unlike typical captions, these contains character names which is very useful. Better to keep it clean, short and convenient description works best. Better not use long and sloppy BS likeA mysteriously enchanting feminine entity of indeterminate yet youthful essence, whose celestial visage radiates with the ethereal luminescence of a thousand dying stars, blessed with locks cascading like the golden rivers of ancient mythology, perhaps styled in a manner reminiscent of contemporary fashion trends though not necessarily adhering to any specific aesthetic paradigm. Her eyes, pools of unfathomable depth and hue, sparkle with the wisdom of millennia yet maintain an innocent quality that defies temporal constraints...For captioning you can use ToriiGate https://huggingface.co/Minthy/ToriiGate-v0.4-7B in short mode.And don't expect it to be as good as flux and others, it tries very hard and after several rolls usually you can get what you want, but it is not that stable and detailed.Lots of Tail/Ears-related concepts:Oh yeahtail censor, holding own tail, hugging own tail, holding another's tail, tail grab, tail raised, tail down, ears down, hand on own ear, tail around own leg, tail around penis, tailjob, tail through clothes, tail under clothes, lifted by tail, tail biting, tail penetration (including a specific indication of vaginal/anal), tail masturbation, holding with tail, panties on tail, bra on tail, tail focus, presenting own tail...(booru meaning, not e621) and many others with natural text. The majority works perfectly, some requires a lot of rolling.Brightness/colors/contrast:You can use extra meta tags to control it:low brightness, high brightness, low saturation, high saturation, low gamma, high gamma, sharp colors, soft colors, hdr, sdrExampleThey work both in epsilon and vpred version and works really good.Epsilon version relies on them too much. Without low brightness or low gamma or limited range (in negative) it might be difficult to achieve true 0,0,0 black, the same often true for white.Both epsilon and vpred versions have like true zsnr, full range of colors and brightness without common flaws observed. But they behaves differently, just try it.Vpred versionMain thing you need to know - lower your CFG from 7 down to 5 (or less). Otherwise, the use is similar with advantages.It seems that starting from v0.7 vpred works flawlessly now. It shouldn't suffer from ignorance of tags close to the 75tokens chunk borders like nai. It is more difficult to get burned images - even on cfg7 usually it just over-saturated but with smooth gradients, which can be useful for some styles. Yes it can make anything from (0,0,0) to (255,255,255). You will find brightness meta tags described above quite useful for easier/lazy prompting, natural text expressions also work. To get the most dark image - put high brightness into negative and/or use low brightness, low gamma tags. If you don't like very bright skin on dark background and want to reduce contrast (or on the contrary, enhance the effect) - use hdr/sdr in negative/positive.It was reported that in rare cases on some prompts there is a drop in contrast. Looks like other vpred models have same behaviour with such prompts, adding a \"separator\" closer to the border of the 75-token chunk fixes this. However, with 0.7 I haven't encountered this myself.To launch vpred version you will need dev build of A1111, Comfy (with special loader node), Forge or Reforge. Just use same parameters (Euler a, cfg 3..5, 20..28 steps) like epsilon. No need to use Cfg rescale, but you can try it, cfg++ works great.Base model:The model here has a small unet polishint after main training to improve small details, bump up resolution and others. Hovewer, you may be also interested into a RouWei-Base, which sometimes can perform better at complex prompts despite having minor mistakes in small details. It also comes in FP32, for example if you want to use fp32 text encoder nodes in Comfy, merge it or finetune.It can be found in Huggingface repo https://huggingface.co/Minthy/RouWei-0.8Known issues:Off course there are:Artists and style tags must be seperated into a different chunk from main prompt or come very lastThere may be some positional or combinational bias in rare cases, but it's not yet clear.There are some complaints about few of the general styles.Epsilon version relies too much on brightness meta tags, sometimes you will need to use them to get desired brightness shiftSome newly added styles/characters might be not as good and disctinct as they deserve toTo be discoveredRequests for artists/characters in future models are open. If you find artist/character/concept that perform weak, inaccurate or has strong watermark - please report, will add them explicitly. Follow for a new versions.JOIN THE DISCORD SERVER https://discord.gg/ZXHENAhqE9License:Same as illustrious. Fell free to use in your merges, finetunes, ets. but please leave a link or mention, it is mandatoryHow it's madeI'll consider to make a report or something like it later. For sure.In short, 98% of work is related to dataset preparations. Instead of blindly relying on loss-weighting based on tag frequency from nai paper, a custom guided loss-weighting implementation along with asynchronous collator for balancing have been used. Ztsnr (or close to it) with Epsilon prediction was achieved using noise scheduler augmentation.Spent compute - over 8k hours of H100 (apart from research and fail attempts)Thanks:First of all I'd like to acknowledge everyone who supports open source, develops in improves code. Thanks to the authors of illustrious for releasing model, thank to NoobAI team for being pioneers in open finetuning of such a scale, sharing experience, raising and solving issues that previously went unnoticed.Personal:Artists wish to remain anonymous for sharing private works; Few anonymous persons - donations, code, captions, etc., Soviet Cat - GPU sponsoring; Sv1. - llm access, captioning, code; K. - training code; Bakariso - datasets, testing, advices, insides; NeuroSenko - donations, testing, code; LOL2024 - a lot of unique datasets; T.,[] - datasets, testing, advises; rred, dga, Fi., ello - donations; TekeshiX - datasets. And other fellow brothers that helped. Love you so much \u2764\ufe0f.And off course everyone who made feedback and requests, it's really valuable.If I forgot to mention anyone, please notify.DonationsIf you want to support - share my models, leave feedback, make a cute picture with kemonomimi-girl. And of course, support original artists.AI is my hobby, I'm spending money on it and not begging for donations. However, it has turned into a large-scale and expensive undertaking. Consider to support to accelerate new training and researches.(Just keep in mind that I can waste it on alcohol or cosplay girls)BTC: bc1qwv83ggq8rvv07uk6dv4njs0j3yygj3aax4wg6cETH/USDT(e): 0x04C8a749F49aE8a56CB84cF0C99CD9E92eDB17dbXMR: 47F7JAyKP8tMBtzwxpoZsUVB8wzg2VrbtDKBice9FAS1FikbHEXXPof4PAb42CQ5ch8p8Hs4RvJuzPHDtaVSdQzD6ZbA5TZif you can offer gpu-time (a100+) - PM.",
+    "sd version": "Other"
+}

Stable-diffusion/rouwei_061Vpred.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ec35d96fc8c6f1b205222077a7d5e929dd2e8d1debb2e130ad2e94e6cb21383
+size 6938040512

Stable-diffusion/rouwei_06Epsilon.html

@@ -0,0 +1,250 @@
+
+                <head>
+                    
+                <meta charset="UTF-8">
+                    <link rel="stylesheet" type="text/css" href="/content/A1111/extensions/sd-civitai-browser-plus/style_html.css">
+                </head>
+                <div class="model-block">
+                    <h2><a href=https://civitai.com/models/950531 target="_blank" id="model_header">RouWei</a></h2>
+                    <h3 class="model-uploader">Uploaded by <a href="https://civitai.com/user/Minthybasis" target="_blank">Minthybasis</a><div class="avatar"><img src=https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/c3116f85-2b3b-4daf-a138-aed342f0cf2d/width=96/Minthybasis.jpeg></div></h3>
+                    <div class="civitai-version-info" style="display:flex; flex-wrap:wrap; justify-content:space-between;">
+                        <dl id="info_block">
+                            <dt>Version</dt>
+                            <dd>0.6 vpred</dd>
+                            <dt>Base Model</dt>
+                            <dd>Illustrious</dd>
+                            <dt>Published</dt>
+                            <dd>2024-11-15</dd>
+                            <dt>Availability</dt>
+                            <dd>Public</dd>
+                            <dt>CivitAI Tags</dt>
+                            <dd>
+                                <div class="civitai-tags-container">
+                                    <span class="civitai-tag">anime</span><span class="civitai-tag">base model</span>
+                                </div>
+                            </dd>
+                            <dt>Download Link</dt>
+                            <dd><a href=https://civitai.com/api/download/models/1064242 target="_blank">https://civitai.com/api/download/models/1064242</a></dd>
+                        </dl>
+                        <div style="align-self:center; min-width:320px;">
+                            <div>
+                                <p style="line-height: 2; font-weight: bold;"><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Use the model without crediting the creator<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Sell images they generate<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Run on services that generate images for money<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Run on Civitai<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Share merges using this model<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="red"><path d="M18 6l-12 12"></path><path d="M6 6l12 12"></path></svg></span> Sell this model or merges using this model<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="red"><path d="M18 6l-12 12"></path><path d="M6 6l12 12"></path></svg></span> Have different permissions when sharing merges</p>
+                            </div>
+                        </div>
+                    </div>
+                    <input type="checkbox" id="preview-civitai-description" class="description-toggle-checkbox">
+                    <div class="model-description">
+                        <h2>Description</h2>
+                        <h2 id="in-depth-retraining-of-illustrious-to-achieve-best-prompt-adherence-knowledge-and-state-of-the-art-performance.-66mudratg">In depth retraining of Illustrious to achieve best prompt adherence, knowledge and state of the art performance.</h2><p><span style="color:rgb(134, 142, 150)">Big dreams come true</span></p><p><span style="color:rgb(134, 142, 150)">The version number is just an index of current final release, not a fraction of the planned training.</span></p><p><a target="_blank" rel="ugc" href="https://huggingface.co/Minthy/RouWei-0.8"><span style="color:rgb(134, 142, 150)">HF repo</span></a></p><p></p><h3 id="large-scale-finetune-using-gpu-cluster-with-a-dataset-of-~13m-pictures-(~4m-with-natural-text-captions)-pqrwpp4f0">Large scale finetune using gpu cluster with a dataset of ~13M pictures (~4M with natural text captions)</h3><ul><li><p>Fresh and wast knowledge about characters, concepts, styles, cultural and related things</p></li><li><p>The best prompt adherence among SDXL anime models at the moment of release</p></li><li><p>Solved main problems with tags bleeding and biases, common for Illustrious, NoobAi and other checkpoints</p></li><li><p>Excellent aesthetics and knowledge across a wide range of styles (<a target="_blank" rel="ugc" href="https://files.catbox.moe/anleso.txt">over 50,000 artists</a> (<a rel="ugc" href="https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg">examples</a>), including hundreds of unique cherry-picked datasets from private galleries, including those received from the artists themselves)</p></li><li><p>High flexibility and variety without stability tradeoff</p></li><li><p>No more annoying watermarks for popular styles thanks to clean dataset</p></li><li><p>Vibrant colors and smooth gradients without trace of burning, full range even with epsilon</p></li><li><p>Pure training from Illustrious v0.1 without involving third-party checkpoints, Loras, tweakers, etc.</p></li></ul><p></p><h3 id="there-are-also-some-issues-and-changes-compared-to-the-previous-version-please-rtfm.-aaa28dbe6">There are also some issues and changes compared to the previous version, please RTFM.</h3><p></p><p>Dataset cut-off - end of April 2025.</p><p></p><h1 id="features-and-prompting:-j89sanl8e">Features and prompting:</h1><p></p><h2 id="important-change:-f6w18r0i0">Important change:</h2><p>When you are prompting artist styles, especially mixing several, <strong><u>their tags MUST BE in a separate CLIP chunk</u></strong>. Just add BREAK after it (for A1111 and derivatives), use conditioning concat node (for Comfy) <strong><u>or at least put them in the very end</u></strong>. Otherwise, significant degradation of results is likely.</p><p></p><h2 id="basic:-aroav4616">Basic:</h2><p></p><p>The checkpoint works both with short-simple and long-complex prompts. However, if there are contradictory or weird things - unlike with others they won't be ignored affecting the output. No guide-rails, no safeguards, no lobotomy.</p><p>Just prompt what you want to see and don't prompt what shouldn't be on the picture. If you want to have a view from above - don't put ceiling into positive, if you want to have crop view with head out of frame - don't make detailed description of character facial features, and so on. Pretty simple but sometimes people are missing it.</p><p></p><p>Version 0.8 comes with advanced understanding of natural text prompts. <u>It doesn't mean that you are obligated to use it</u>, tags only - completely fine, especially because understanding of tags combinations is also improved.</p><p>Do not expect it to perform like Flux or other models based on T5 or LLM text encoders. The whole size ot SDXL checkpoint is less then only that text encoder, in addition illustrious-v0.1 which is used as the base completely forgot a lot of general things from vanilla sdxl-base.</p><p>However, even in current state it works much better, allows to do new things usually impossible without external guidance, as well making manual editing, inpainting, etc more convenient.</p><p></p><p>To achieve best performance you should keep track of CLIP chunks. In SDXL the prompt is separated into a chunks of 75 (77 including BOS and EOS) tokens, that are processing by CLIP separately, and only then are concatinating and comes as conditions to unet.</p><p>If you want to specify some features for character/object and separate them from other prompt parts - make sure they are in the same chunk and optionally separate it with <code style="text-wrap: wrap">BREAK</code>. It will not solve problem of traits mixing completely, but can reduce it improving overall understanding, since text encoders on RouWei are able to process the whole sequence, not individual concepts better then others.</p><p>Dataset contains only booru-style tags and natural text expressions. Despite having a share of furries, real life photos, western media, etc. all captions have been converted to classic booru style to avoid a number of problems from mixing of different systems. So e621 tags won't be understanded properly.</p><p></p><h3 id="sampling-parameters:-7andwk13p">Sampling parameters:</h3><ul><li><p>~1 megapixel for txt2img, any AR with resolution multiple of 32 (1024x1024, 1056x, 1152x, 1216x832,...). Euler_a, 20..28steps.</p></li><li><p>CFG: for epsilon version 4..9 (7 is best), for vpred version, 3..5</p></li><li><p>Sigmas multiply may improve results a bit, CFG++ samplers work fine. LCM/PCM/DMD/... and exotic samplers untested.</p></li><li><p><strong>Some schedulers doesn't work well.</strong></p></li><li><p>Highresfix - x1.5 latent + denoise 0.6 or any gan + denoise 0.3..0.55.</p></li><li><p>For vpred version lower CFG 3..5 is needed!</p></li></ul><p></p><h3 id="for-vpred-version-lower-cfg-3..5-is-needed!-uwz9s21js"><strong>For vpred version lower CFG 3..5 is needed!</strong></h3><p></p><h2 id="quality-classification:-kw87tas72">Quality classification:</h2><p>Only 4 quality tags:</p><pre><code style="text-wrap: wrap">masterpiece, best quality</code></pre><p>for positive and</p><pre><code style="text-wrap: wrap">low quality, worst quality</code></pre><p>for negative.</p><p>Nothing else. Actually you can even omit positive and reduce negative to <code style="text-wrap: wrap">low quality</code> only, since they can affect basic style and composition.</p><p>Meta tags like <code style="text-wrap: wrap">lowres</code> have been removed and don't work, better not to use them. Low resolution images have been either removed or upscaled and cleaned with DAT depending on their importance.</p><p></p><h2 id="negative-prompt:-79eot1hhx">Negative prompt:</h2><pre><code style="text-wrap: wrap">worst quality, low quality, watermark</code></pre><p>That's all, no need of "rusty trombone", "farting on prey" and others. Do not put tags like greyscale, monochrome in negative unless you understand what are you doing. Extra tags for brightness/colors/contrast section below can be used</p><p></p><h2 id="artist-styles:-6yumq5hap">Artist styles:</h2><p></p><p><a target="_blank" rel="ugc" href="https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg">Grids with examples</a>, <a target="_blank" rel="ugc" href="https://files.catbox.moe/anleso.txt">list/wildcard</a> (also can be found in "training data").</p><p></p><h3 id="used-with-&quot;by-&quot;-it's-mandatory.-it-will-not-work-properly-without-it.-innc8cnap"><u>Used with "by " it's mandatory</u>. It will not work properly without it.</h3><p>"by " is a meta-token for styles to avoid mixing/misinterpret with tags/characters of similar or close name. This allows to have a better results for styles and at the same time avoid random style fluctuation that you may observe in other checkpoints.</p><p></p><p>Multiple give very interesting results, can be controlled with prompt weights and spells.</p><p></p><h3 id="you-must-add-break-after-artistsstyle-tags-(for-a1111)-or-concat-conditioning-(for-comfy)-or-put-them-in-the-very-end-of-your-prompt.-36i67w9c0">YOU MUST ADD <code style="text-wrap: wrap">BREAK</code> after artists/style tags (for A1111) or concat conditioning (for Comfy) or put them in the very end of your prompt.</h3><p>For example:</p><pre><code style="text-wrap: wrap">by kantoku, by wlop, best quality, masterpiece BREAK 1girl, ...</code></pre><p></p><h2 id="general-styles:-5ig9sgqza">General styles:</h2><pre><code style="text-wrap: wrap">2.5d, anime screencap, bold line, sketch, cgi, digital painting, flat colors, smooth shading, minimalistic, ink style, oil style, pastel style</code></pre><h3 id="booru-tags-styles:-uq05a7u5y">Booru tags styles:</h3><pre><code style="text-wrap: wrap">1950s (style), 1960s (style), 1970s (style), 1980s (style), 1990s (style), 2000s (style), animification, art nouveau, pinup (style), toon (style), western comics (style), nihonga, shikishi, minimalism, fine art parody</code></pre><p>and everything from <a target="_blank" rel="ugc" href="https://danbooru.donmai.us/wiki_pages/traditional_media">this group</a>.</p><p>Can be used in combinations (with artists too), with weights, both in positive and negative prompts.</p><p></p><h2 id="characters:-8htyrng9m">Characters:</h2><p>Use full name booru tag and proper formatting, like <code style="text-wrap: wrap">karin_(blue_archive)</code> -&gt; <code style="text-wrap: wrap">karin \(blue archive\)</code>, use skin tags for better reproducing, like <code style="text-wrap: wrap">karin \(bunny\) \(blue archive\)</code>. Autocomplete extension might be very useful.</p><p>Most characters are recognized just by their booru tag, but it will be more accurate if you describe their basic traits. Here you can easily redress your waifu/husbendo just by the prompt without suffering from the typical leaks of basic features.</p><p></p><h2 id="natural-text:-cq8whprlj">Natural text:</h2><p>Use it in combination with booru tags, works great. Use only natural text after typing styles and quality tags. Use just booru tags and forget about it, it's all up to you. To get best performance keep track if CLIP 75 tokens chunks.</p><p></p><p>About 4M of images in dataset had hybrid natural-text captions, made by Claude, GPT, Gemini, <a target="_blank" rel="ugc" href="https://huggingface.co/Minthy/ToriiGate-v0.4-7B">ToriiGate</a>, then refactored, cleaned and combined with tags in different variations for augmentation.</p><p></p><p>Unlike typical captions, these contains character names which is very useful. Better to keep it clean, short and convenient description works best. Better <strong><u>not</u></strong> use long and sloppy BS like</p><pre><code style="text-wrap: wrap">A mysteriously enchanting feminine entity of indeterminate yet youthful essence, whose celestial visage radiates with the ethereal luminescence of a thousand dying stars, blessed with locks cascading like the golden rivers of ancient mythology, perhaps styled in a manner reminiscent of contemporary fashion trends though not necessarily adhering to any specific aesthetic paradigm. Her eyes, pools of unfathomable depth and hue, sparkle with the wisdom of millennia yet maintain an innocent quality that defies temporal constraints...</code></pre><p>For captioning you can use <a target="_blank" rel="ugc" href="https://huggingface.co/Minthy/ToriiGate-v0.4-7B">ToriiGate</a> in short mode.</p><p></p><p>And don't expect it to be as good as flux and others, it tries very hard and after several rolls usually you can get what you want, but it is not that stable and detailed.</p><p></p><h2 id="lots-of-tailears-related-concepts:-qsxmk9gzm">Lots of Tail/Ears-related concepts:</h2><p>Oh yeah</p><pre><code style="text-wrap: wrap">tail censor, holding own tail, hugging own tail, holding another's tail, tail grab, tail raised, tail down, ears down, hand on own ear, tail around own leg, tail around penis, tailjob, tail through clothes, tail under clothes, lifted by tail, tail biting, tail penetration (including a specific indication of vaginal/anal), tail masturbation, holding with tail, panties on tail, bra on tail, tail focus, presenting own tail...</code></pre><p>(booru meaning, not e621) and many others with natural text. The majority works perfectly, some requires a lot of rolling.</p><p></p><h1 id="brightnesscolorscontrast:-lk2g7nwtj">Brightness/colors/contrast:</h1><p>You can use extra meta tags to control it:</p><pre><code style="text-wrap: wrap">low brightness, high brightness, low saturation, high saturation, low gamma, high gamma, sharp colors, soft colors, hdr, sdr</code></pre><p><a target="_blank" rel="ugc" href="https://files.catbox.moe/r8wrb6.jpg">Example</a></p><p>They work both in epsilon and vpred version and works really good.</p><p>Epsilon version relies on them too much. Without low brightness or low gamma or limited range (in negative) it might be difficult to achieve true 0,0,0 black, the same often true for white.</p><p>Both epsilon and vpred versions have like true zsnr, full range of colors and brightness without common flaws observed. But they behaves differently, just try it.</p><p></p><h1 id="vpred-version-9pr2w5h4d">Vpred version</h1><p></p><p>Main thing you need to know - lower your CFG from 7 down to 5 (or less). Otherwise, the use is similar with advantages.</p><p>It seems that starting from v0.7 vpred works flawlessly now. It shouldn't suffer from ignorance of tags close to the 75tokens chunk borders like nai. It is more difficult to get burned images - even on cfg7 usually it just over-saturated but with smooth gradients, which can be useful for some styles. Yes it can make anything from (0,0,0) to (255,255,255). You will find brightness meta tags described above quite useful for easier/lazy prompting, natural text expressions also work. To get the most dark image - put <code style="text-wrap: wrap">high brightness</code> into <u>negative</u> and/or use <code style="text-wrap: wrap">low brightness, low gamma</code> tags. If you don't like very bright skin on dark background and want to reduce contrast (or on the contrary, enhance the effect) - use hdr/sdr in negative/positive.</p><p>It was reported that in rare cases on some prompts there is a drop in contrast. Looks like other vpred models have same behaviour with such prompts, adding a "separator" closer to the border of the 75-token chunk fixes this. However, with 0.7 I haven't encountered this myself.</p><p></p><p><span style="color:rgb(134, 142, 150)">To launch vpred version you will need dev build of A1111, Comfy (with special loader node), Forge or Reforge. Just use same parameters (Euler a, cfg 3..5, 20..28 steps) like epsilon. No need to use Cfg rescale, but you can try it, cfg++ works great.</span></p><p></p><h1 id="base-model:-lwa0dfdbf">Base model:</h1><p>The model here has a small unet polishint after main training to improve small details, bump up resolution and others. Hovewer, you may be also interested into a RouWei-Base, which sometimes can perform better at complex prompts despite having minor mistakes in small details. It also comes in FP32, for example if you want to use fp32 text encoder nodes in Comfy, merge it or finetune.</p><p>It can be found in <a target="_blank" rel="ugc" href="https://huggingface.co/Minthy/RouWei-0.8">Huggingface repo</a></p><p></p><h1 id="known-issues:-jxgt6yemu">Known issues:</h1><p>Off course there are:</p><ul><li><p>Artists and style tags must be seperated into a different chunk from main prompt or come very last</p></li><li><p>There may be some positional or combinational bias in rare cases, but it's not yet clear.</p></li><li><p>There are some complaints about few of the general styles.</p></li><li><p>Epsilon version relies too much on brightness meta tags, sometimes you will need to use them to get desired brightness shift</p></li><li><p>Some newly added styles/characters might be not as good and disctinct as they deserve to</p></li><li><p>To be discovered</p><p></p></li><li><p></p></li></ul><h3 id="requests-for-artistscharacters-in-future-models-are-open.-if-you-find-artistcharacterconcept-that-perform-weak-inaccurate-or-has-strong-watermark-please-report-will-add-them-explicitly.-follow-for-a-new-versions.-kgc39ed4e">Requests for artists/characters in future models are open. If you find artist/character/concept that perform weak, inaccurate or has strong watermark - please report, will add them explicitly. Follow for a new versions.</h3><p></p><h2 id="join-the-discord-server-9lpaojolv"><a target="_blank" rel="ugc" href="https://discord.gg/ZXHENAhqE9">JOIN THE DISCORD SERVER</a></h2><p></p><h2 id="license:-0hpy342uh">License:</h2><p>Same as illustrious. Fell free to use in your merges, finetunes, ets. but please leave a link or mention, it is mandatory</p><p></p><h2 id="how-it's-made-zc68i8qve">How it's made</h2><p>I'll consider to make a report or something like it later. For sure.</p><p>In short, 98% of work is related to dataset preparations. Instead of blindly relying on loss-weighting based on tag frequency from nai paper, a custom guided loss-weighting implementation along with asynchronous collator for balancing have been used. Ztsnr (or close to it) with Epsilon prediction was achieved using noise scheduler augmentation.</p><p></p><p>Spent compute - over 8k hours of H100 (apart from research and fail attempts)</p><p></p><h1 id="thanks:-4tsy2pj9q">Thanks:</h1><p>First of all I'd like to acknowledge everyone who supports open source, develops in improves code. Thanks to the authors of illustrious for releasing model, thank to NoobAI team for being pioneers in open finetuning of such a scale, sharing experience, raising and solving issues that previously went unnoticed.</p><h3 id="personal:-9a91vqx2i">Personal:</h3><p>Artists wish to remain anonymous for sharing private works; Few anonymous persons - donations, code, captions, etc., Soviet Cat - GPU sponsoring; Sv1. - llm access, captioning, code; K. - training code; Bakariso - datasets, testing, advices, insides; NeuroSenko - donations, testing, code; LOL2024 - a lot of unique datasets; T.,[] - datasets, testing, advises; rred, dga, Fi., ello - donations; TekeshiX - datasets. And other fellow brothers that helped. Love you so much ❤️.</p><p>And off course everyone who made feedback and requests, it's really valuable.</p><p><span style="color:rgb(134, 142, 150)">If I forgot to mention anyone, please notify.</span></p><p></p><h2 id="donations-jdo27wd0z">Donations</h2><p>If you want to support - share my models, leave feedback, make a cute picture with kemonomimi-girl. And of course, support original artists.</p><p>AI is my hobby, I'm spending money on it and not begging for donations. However, it has turned into a large-scale and expensive undertaking. Consider to support to accelerate new training and researches.</p><p><span style="color:rgb(134, 142, 150)">(Just keep in mind that I can waste it on alcohol or cosplay girls)</span></p><p>BTC: bc1qwv83ggq8rvv07uk6dv4njs0j3yygj3aax4wg6c</p><p>ETH/USDT(e): 0x04C8a749F49aE8a56CB84cF0C99CD9E92eDB17db</p><p>XMR: 47F7JAyKP8tMBtzwxpoZsUVB8wzg2VrbtDKBice9FAS1FikbHEXXPof4PAb42CQ5ch8p8Hs4RvJuzPHDtaVSdQzD6ZbA5TZ</p><p>if you can offer gpu-time (a100+) - PM.</p><p></p><p></p><p></p><p></p><p></p>
+                    </div>
+                    <label for="preview-civitai-description" class="description-toggle-label"></label>
+                </div>
+                <div align=center><div class="sampleimgs"><input type="radio" name="zoomRadio" id="resetZoom" class="zoom-radio" checked>
+                    <div class="model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_0" class="zoom-radio">
+                    <label for="zoomRadiopreview_0" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/f96c7ce9-0eaf-4461-bee3-176a008ee8d2/width=1536/40258578.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/f96c7ce9-0eaf-4461-bee3-176a008ee8d2/width=1536/40258578.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>masterpiece, best quality, good quality,
+1girl, konpaku youmu, half-closed eyes, solo, black hairband, (konpaku youmu (ghost):1.2), hairband, green vest, puffy sleeves, upper body, bow, black bowtie, black bow, short hair, vest, bowtie, ribbon, leaf, shirt, wind, short sleeves, black background, looking down, white shirt, hitodama, black ribbon, hair ribbon, puffy short sleeves, grey hair, simple background, floating hair, closed mouth, expressionless, wing collar, hitodama print, buttons, green eyes, collared shirt, falling leaves, aqua eyes, glowing eyes, pale skin, (skeleton, split screen, straight-on, android, crank, gauge, gears, glowing, light bulb, mechabare, sparkle, x-ray:1.2), 
+&lt;lora:youmuanon_noob-vpred0.1_flawless_ds_v2:1&gt;</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, bad quality, lowres, jpeg artifacts, bad anatomy, bad hands, multiple views, signature, watermark, censored, covered navel, limited range, high brightness, high contrast,</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>4147779862</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1024x1024</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred_rc1</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>5</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_0">
+                                    <label class="tab-label" for="chckpreview_0">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl-vae-fp16-fix.safetensors</dd></div><div class="civitai-meta"><dt>Ensd</dt><dd>31337</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;62c7c729ad&#x27;, &#x27;model&#x27;: &#x27;e63d568c63&#x27;, &#x27;lora:youmuanon_noob-vpred0.1_flawless_ds_v2&#x27;: &#x27;7b7e860a9bae&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.3.6dev2-v1.10.1RC-latest-1453-g0fd79513</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;7b7e860a9bae&#x27;, &#x27;name&#x27;: &#x27;youmuanon_noob-vpred0.1_flawless_ds_v2&#x27;, &#x27;type&#x27;: &#x27;lora&#x27;, &#x27;weight&#x27;: 1}, {&#x27;hash&#x27;: &#x27;e63d568c63&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred_rc1&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>e63d568c63</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Align Your Steps</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>2x-AnimeSharpV3</dd></div><div class="civitai-meta"><dt>Adetailer model</dt><dd>face_yolov8n.pt</dd></div><div class="civitai-meta"><dt>Adetailer version</dt><dd>24.9.0</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Sigmas_merge_mode</dt><dd>Multiply</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.5</dd></div><div class="civitai-meta"><dt>Sigmas_mult_factor</dt><dd>2</dd></div><div class="civitai-meta"><dt>Adetailer mask blur</dt><dd>8</dd></div><div class="civitai-meta"><dt>Adetailer confidence</dt><dd>0.3</dd></div><div class="civitai-meta"><dt>Sigmas_merge_enabled</dt><dd>True</dd></div><div class="civitai-meta"><dt>Adetailer dilate erode</dt><dd>4</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Adetailer inpaint padding</dt><dd>32</dd></div><div class="civitai-meta"><dt>Adetailer negative prompt</dt><dd>{}</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div><div class="civitai-meta"><dt>Adetailer denoising strength</dt><dd>0.4</dd></div><div class="civitai-meta"><dt>Old prompt editing timelines</dt><dd>True</dd></div><div class="civitai-meta"><dt>Adetailer inpaint only masked</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_1" class="zoom-radio">
+                    <label for="zoomRadiopreview_1" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/76198546-76cd-4bf4-9fc8-f935da633bbf/width=1328/40258585.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/76198546-76cd-4bf4-9fc8-f935da633bbf/width=1328/40258585.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>(by deel \(rkeg\):0.8), best quality, masterpiece,
+1girl, folinic \(arknights\), cat girl, brown hair, multicolored hair, cat ears, surprised, blush, hair flower, holding papers, BREAK
+indoors, scientific lab, glass floor, intricate interior, computers, research facility</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, watermark</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>2290774044</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>832x1216</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_1">
+                                    <label class="tab-label" for="chckpreview_1">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;84c8888446&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;84c8888446&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>84c8888446</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.6</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_2" class="zoom-radio">
+                    <label for="zoomRadiopreview_2" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/5033de43-81ec-418e-9429-f2a7425353c4/width=1728/40258592.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/5033de43-81ec-418e-9429-f2a7425353c4/width=1728/40258592.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by ratatatat74, by tarakanovich,cube85, masterpiece, best quality,
+1girl, fake animal ears, cat ears, long black hair, large breasts, erect nipples, completely nude, butt plug, black thighhighs, garter straps, squatting, reverse cowgirl position, girl on top, from behind, ass, mature female, pussy juice, ass grab, large penis, vaginal, pov hands, tally, (slap mark:0.7),
+indoors, bedroom, dark night, dimly lit room, dark blue theme, low brightness, low gamma, sdr</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, limited range</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>1178610355</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1152x896</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_2">
+                                    <label class="tab-label" for="chckpreview_2">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_3" class="zoom-radio">
+                    <label for="zoomRadiopreview_3" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/9af10cd7-4513-4147-a4f2-033f46afd3e0/width=1824/40258577.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/9af10cd7-4513-4147-a4f2-033f46afd3e0/width=1824/40258577.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>(by ratatatat74:0.7), by tarakanovich, cube85, masterpiece, best quality,
+1girl, vulpisfoglia \(arknights\), fox girl, fox ears, large ears, ears down, blonde hair, large breasts, yellow eyes, fox tail, nightgown, see-through, bottomless, from behind, tail censor, (scar on animal ear:0.8), white panties, naughty face,
+all fours, fox shadow puppet, glowing eyes,
+black background, low brightness, low gamma</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, extra ears</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>1556719420</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1216x832</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>20</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_3">
+                                    <label class="tab-label" for="chckpreview_3">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_4" class="zoom-radio">
+                    <label for="zoomRadiopreview_4" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/0d48a8dd-f3d3-42f6-b79d-de7de7b52139/width=1200/40258590.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/0d48a8dd-f3d3-42f6-b79d-de7de7b52139/width=1200/40258590.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>masterpiece, best quality, by merrytail, by sub-res, by kie \(wylee2212\),
+1girl, fox devil \(chainsaw man\), fox girl, fox ears, fox tail, whisker markings, long white hair, red eyes, (extra eyes:1.3), fangs, evil smile, white kimono with floral print, collarbone, hair bun, fox shadow puppet, cowboy shot, posing,
+black background, low brightness</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, limited range, monster, 1other, animal, fox</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>1376750140</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>800x1344</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_4">
+                                    <label class="tab-label" for="chckpreview_4">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_5" class="zoom-radio">
+                    <label for="zoomRadiopreview_5" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/209ff688-e01d-41d4-b842-a3ffadd3741f/width=1824/40258575.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/209ff688-e01d-41d4-b842-a3ffadd3741f/width=1824/40258575.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>masterpiece, best quality, by done \(donezumi\),
+1girl, kon \(tokyo ravens\), tokyo ravens, fox girl, fox ears, fox tail, white hair, thick eyebrows, pale skin, blue eyes, white aqua japanese clothes, detached sleeves, ribbon trim, fox shadow puppet, pom pom \(clothes\), black shorts, white thighhighs, hips, happy, BREAK floating in air, legs apart, posing, full body, wide shot,
+white background, high brightness</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>2168859150</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1216x832</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>20</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_5">
+                                    <label class="tab-label" for="chckpreview_5">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;84c8888446&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;84c8888446&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>84c8888446</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_6" class="zoom-radio">
+                    <label for="zoomRadiopreview_6" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/90435b3a-ef7d-4be2-b700-00601088c553/width=1248/40258594.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/90435b3a-ef7d-4be2-b700-00601088c553/width=1248/40258594.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by tandohark, (by personal ami:0.6), best quality, masterpiece,
+2girls, asymmetrical docking BREAK
+izuna \(blue archive\), fox girl, fox ears, fox mask, fox tail, halo, topless, denim shorts, striped bikini bra, medium breasts, sideboob, visors cap BREAK
+wakamo \(blue archive\), fox girl, fox ears, fox tail, long black hair, red halo, hair ornament, large breasts, fox mask, white bikini swimsuit, collarbone, :3 BREAK
+outdoors, beach, seashore, palms, clear blue sky</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, watermark</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>1114850771</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>832x1216</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>5</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_6">
+                                    <label class="tab-label" for="chckpreview_6">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;84c8888446&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;84c8888446&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>84c8888446</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_7" class="zoom-radio">
+                    <label for="zoomRadiopreview_7" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/23c627f5-451b-47de-a1d3-7dce53adb8d9/width=1248/40258589.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/23c627f5-451b-47de-a1d3-7dce53adb8d9/width=1248/40258589.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by akai sashimi, by nat the lich, by foxykuro, by nekoya \(liu\), masterpiece,  best quality,
+1girl, kal&#x27;tsit \(arknights\), cat girl, lynx ears, silver hair, green eyes/, naked towel, covered nipples, off shoulder, oripathy lesion \(arknights\), collarbone, shaded face, (evil smile:1.1), smug, open mouth, angry, fang, tilted head, wet,
+holding handcuffs,
+cowboy shot, motion lines,
+indoors, office, dimmed, dark background</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, watermark, dutch angle</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>2542436241</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>832x1216</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_7">
+                                    <label class="tab-label" for="chckpreview_7">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Ensd</dt><dd>31337</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_8" class="zoom-radio">
+                    <label for="zoomRadiopreview_8" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/fd31b937-db26-404f-ba61-c86eecedb37c/width=1328/40258587.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/fd31b937-db26-404f-ba61-c86eecedb37c/width=1328/40258587.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by akai sashimi, by nat the lich, by foxykuro, by nekoya \(liu\), masterpiece, best quality,
+1girl, kal&#x27;tsit \(arknights\), cat girl, lynx ears, silver hair, green eyes, short tail, completely nude, oripathy lesion \(arknights\), scars on spine, reverse cowgirl position, motion lines, from below, open mouth, smug, heavenly ass, v,
+1boy, vaginal, sex, pov, deep penetration, pussy juice trail,
+indoors, bedroom, dimmed, dark background</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, watermark, abstract background</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>3956367149</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>832x1216</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>20</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_8">
+                                    <label class="tab-label" for="chckpreview_8">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.6</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_9" class="zoom-radio">
+                    <label for="zoomRadiopreview_9" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/11c49fc0-f6fa-40b2-bb86-30e3a2c9cd0a/width=2064/40258599.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/11c49fc0-f6fa-40b2-bb86-30e3a2c9cd0a/width=2064/40258599.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by kate fox, masterpiece, best quality,
+1girl, yuel \(granblue fantasy\), fox girl, yellow eyes, fox ears, large fox tail, bikini bra, fur trim, elbow gloves, :3, wide shot, BREAK
+outdoors, torii, asian nature, epic landscape of mountains, above clouds</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, bad hands, watermark
+worst quality, low quality</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>2135703813</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1376x720</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>20</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_9">
+                                    <label class="tab-label" for="chckpreview_9">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;84c8888446&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;84c8888446&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>84c8888446</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Rescale_cfg_enabled</dt><dd>True</dd></div><div class="civitai-meta"><dt>Rescale_cfg_multiplier</dt><dd>0.7</dd></div></div></div></div></dl></div></div></div></div>
+                

Stable-diffusion/rouwei_06Epsilon.json

@@ -0,0 +1,8 @@
+{
+    "sha256": "B96CD67807DD06F42315905BF4E50D5B9A1EC9D6FECA1ADB40F3A35FE8BA5F00",
+    "modelId": 950531,
+    "modelVersionId": 1064213,
+    "activation text": "",
+    "description": "In depth retraining of Illustrious to achieve best prompt adherence, knowledge and state of the art performance.Big dreams come trueThe version number is just an index of current final release, not a fraction of the planned training.HF repo https://huggingface.co/Minthy/RouWei-0.8Large scale finetune using gpu cluster with a dataset of ~13M pictures (~4M with natural text captions)Fresh and wast knowledge about characters, concepts, styles, cultural and related thingsThe best prompt adherence among SDXL anime models at the moment of releaseSolved main problems with tags bleeding and biases, common for Illustrious, NoobAi and other checkpointsExcellent aesthetics and knowledge across a wide range of styles (over 50,000 artists https://files.catbox.moe/anleso.txt (examples https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg), including hundreds of unique cherry-picked datasets from private galleries, including those received from the artists themselves)High flexibility and variety without stability tradeoffNo more annoying watermarks for popular styles thanks to clean datasetVibrant colors and smooth gradients without trace of burning, full range even with epsilonPure training from Illustrious v0.1 without involving third-party checkpoints, Loras, tweakers, etc.There are also some issues and changes compared to the previous version, please RTFM.Dataset cut-off - end of April 2025.Features and prompting:Important change:When you are prompting artist styles, especially mixing several, their tags MUST BE in a separate CLIP chunk. Just add BREAK after it (for A1111 and derivatives), use conditioning concat node (for Comfy) or at least put them in the very end. Otherwise, significant degradation of results is likely.Basic:The checkpoint works both with short-simple and long-complex prompts. However, if there are contradictory or weird things - unlike with others they won't be ignored affecting the output. No guide-rails, no safeguards, no lobotomy.Just prompt what you want to see and don't prompt what shouldn't be on the picture. If you want to have a view from above - don't put ceiling into positive, if you want to have crop view with head out of frame - don't make detailed description of character facial features, and so on. Pretty simple but sometimes people are missing it.Version 0.8 comes with advanced understanding of natural text prompts. It doesn't mean that you are obligated to use it, tags only - completely fine, especially because understanding of tags combinations is also improved.Do not expect it to perform like Flux or other models based on T5 or LLM text encoders. The whole size ot SDXL checkpoint is less then only that text encoder, in addition illustrious-v0.1 which is used as the base completely forgot a lot of general things from vanilla sdxl-base.However, even in current state it works much better, allows to do new things usually impossible without external guidance, as well making manual editing, inpainting, etc more convenient.To achieve best performance you should keep track of CLIP chunks. In SDXL the prompt is separated into a chunks of 75 (77 including BOS and EOS) tokens, that are processing by CLIP separately, and only then are concatinating and comes as conditions to unet.If you want to specify some features for character/object and separate them from other prompt parts - make sure they are in the same chunk and optionally separate it with BREAK. It will not solve problem of traits mixing completely, but can reduce it improving overall understanding, since text encoders on RouWei are able to process the whole sequence, not individual concepts better then others.Dataset contains only booru-style tags and natural text expressions. Despite having a share of furries, real life photos, western media, etc. all captions have been converted to classic booru style to avoid a number of problems from mixing of different systems. So e621 tags won't be understanded properly.Sampling parameters:~1 megapixel for txt2img, any AR with resolution multiple of 32 (1024x1024, 1056x, 1152x, 1216x832,...). Euler_a, 20..28steps.CFG: for epsilon version 4..9 (7 is best), for vpred version, 3..5Sigmas multiply may improve results a bit, CFG++ samplers work fine. LCM/PCM/DMD/... and exotic samplers untested.Some schedulers doesn't work well.Highresfix - x1.5 latent + denoise 0.6 or any gan + denoise 0.3..0.55.For vpred version lower CFG 3..5 is needed!For vpred version lower CFG 3..5 is needed!Quality classification:Only 4 quality tags:masterpiece, best qualityfor positive andlow quality, worst qualityfor negative.Nothing else. Actually you can even omit positive and reduce negative to low quality only, since they can affect basic style and composition.Meta tags like lowres have been removed and don't work, better not to use them. Low resolution images have been either removed or upscaled and cleaned with DAT depending on their importance.Negative prompt:worst quality, low quality, watermarkThat's all, no need of \"rusty trombone\", \"farting on prey\" and others. Do not put tags like greyscale, monochrome in negative unless you understand what are you doing. Extra tags for brightness/colors/contrast section below can be usedArtist styles:Grids with examples https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg, list/wildcard https://files.catbox.moe/anleso.txt (also can be found in \"training data\").Used with \"by \" it's mandatory. It will not work properly without it.\"by \" is a meta-token for styles to avoid mixing/misinterpret with tags/characters of similar or close name. This allows to have a better results for styles and at the same time avoid random style fluctuation that you may observe in other checkpoints.Multiple give very interesting results, can be controlled with prompt weights and spells.YOU MUST ADD BREAK after artists/style tags (for A1111) or concat conditioning (for Comfy) or put them in the very end of your prompt.For example:by kantoku, by wlop, best quality, masterpiece BREAK 1girl, ...General styles:2.5d, anime screencap, bold line, sketch, cgi, digital painting, flat colors, smooth shading, minimalistic, ink style, oil style, pastel styleBooru tags styles:1950s (style), 1960s (style), 1970s (style), 1980s (style), 1990s (style), 2000s (style), animification, art nouveau, pinup (style), toon (style), western comics (style), nihonga, shikishi, minimalism, fine art parodyand everything from this group https://danbooru.donmai.us/wiki_pages/traditional_media.Can be used in combinations (with artists too), with weights, both in positive and negative prompts.Characters:Use full name booru tag and proper formatting, like karin_(blue_archive) -> karin \\(blue archive\\), use skin tags for better reproducing, like karin \\(bunny\\) \\(blue archive\\). Autocomplete extension might be very useful.Most characters are recognized just by their booru tag, but it will be more accurate if you describe their basic traits. Here you can easily redress your waifu/husbendo just by the prompt without suffering from the typical leaks of basic features.Natural text:Use it in combination with booru tags, works great. Use only natural text after typing styles and quality tags. Use just booru tags and forget about it, it's all up to you. To get best performance keep track if CLIP 75 tokens chunks.About 4M of images in dataset had hybrid natural-text captions, made by Claude, GPT, Gemini, ToriiGate https://huggingface.co/Minthy/ToriiGate-v0.4-7B, then refactored, cleaned and combined with tags in different variations for augmentation.Unlike typical captions, these contains character names which is very useful. Better to keep it clean, short and convenient description works best. Better not use long and sloppy BS likeA mysteriously enchanting feminine entity of indeterminate yet youthful essence, whose celestial visage radiates with the ethereal luminescence of a thousand dying stars, blessed with locks cascading like the golden rivers of ancient mythology, perhaps styled in a manner reminiscent of contemporary fashion trends though not necessarily adhering to any specific aesthetic paradigm. Her eyes, pools of unfathomable depth and hue, sparkle with the wisdom of millennia yet maintain an innocent quality that defies temporal constraints...For captioning you can use ToriiGate https://huggingface.co/Minthy/ToriiGate-v0.4-7B in short mode.And don't expect it to be as good as flux and others, it tries very hard and after several rolls usually you can get what you want, but it is not that stable and detailed.Lots of Tail/Ears-related concepts:Oh yeahtail censor, holding own tail, hugging own tail, holding another's tail, tail grab, tail raised, tail down, ears down, hand on own ear, tail around own leg, tail around penis, tailjob, tail through clothes, tail under clothes, lifted by tail, tail biting, tail penetration (including a specific indication of vaginal/anal), tail masturbation, holding with tail, panties on tail, bra on tail, tail focus, presenting own tail...(booru meaning, not e621) and many others with natural text. The majority works perfectly, some requires a lot of rolling.Brightness/colors/contrast:You can use extra meta tags to control it:low brightness, high brightness, low saturation, high saturation, low gamma, high gamma, sharp colors, soft colors, hdr, sdrExampleThey work both in epsilon and vpred version and works really good.Epsilon version relies on them too much. Without low brightness or low gamma or limited range (in negative) it might be difficult to achieve true 0,0,0 black, the same often true for white.Both epsilon and vpred versions have like true zsnr, full range of colors and brightness without common flaws observed. But they behaves differently, just try it.Vpred versionMain thing you need to know - lower your CFG from 7 down to 5 (or less). Otherwise, the use is similar with advantages.It seems that starting from v0.7 vpred works flawlessly now. It shouldn't suffer from ignorance of tags close to the 75tokens chunk borders like nai. It is more difficult to get burned images - even on cfg7 usually it just over-saturated but with smooth gradients, which can be useful for some styles. Yes it can make anything from (0,0,0) to (255,255,255). You will find brightness meta tags described above quite useful for easier/lazy prompting, natural text expressions also work. To get the most dark image - put high brightness into negative and/or use low brightness, low gamma tags. If you don't like very bright skin on dark background and want to reduce contrast (or on the contrary, enhance the effect) - use hdr/sdr in negative/positive.It was reported that in rare cases on some prompts there is a drop in contrast. Looks like other vpred models have same behaviour with such prompts, adding a \"separator\" closer to the border of the 75-token chunk fixes this. However, with 0.7 I haven't encountered this myself.To launch vpred version you will need dev build of A1111, Comfy (with special loader node), Forge or Reforge. Just use same parameters (Euler a, cfg 3..5, 20..28 steps) like epsilon. No need to use Cfg rescale, but you can try it, cfg++ works great.Base model:The model here has a small unet polishint after main training to improve small details, bump up resolution and others. Hovewer, you may be also interested into a RouWei-Base, which sometimes can perform better at complex prompts despite having minor mistakes in small details. It also comes in FP32, for example if you want to use fp32 text encoder nodes in Comfy, merge it or finetune.It can be found in Huggingface repo https://huggingface.co/Minthy/RouWei-0.8Known issues:Off course there are:Artists and style tags must be seperated into a different chunk from main prompt or come very lastThere may be some positional or combinational bias in rare cases, but it's not yet clear.There are some complaints about few of the general styles.Epsilon version relies too much on brightness meta tags, sometimes you will need to use them to get desired brightness shiftSome newly added styles/characters might be not as good and disctinct as they deserve toTo be discoveredRequests for artists/characters in future models are open. If you find artist/character/concept that perform weak, inaccurate or has strong watermark - please report, will add them explicitly. Follow for a new versions.JOIN THE DISCORD SERVER https://discord.gg/ZXHENAhqE9License:Same as illustrious. Fell free to use in your merges, finetunes, ets. but please leave a link or mention, it is mandatoryHow it's madeI'll consider to make a report or something like it later. For sure.In short, 98% of work is related to dataset preparations. Instead of blindly relying on loss-weighting based on tag frequency from nai paper, a custom guided loss-weighting implementation along with asynchronous collator for balancing have been used. Ztsnr (or close to it) with Epsilon prediction was achieved using noise scheduler augmentation.Spent compute - over 8k hours of H100 (apart from research and fail attempts)Thanks:First of all I'd like to acknowledge everyone who supports open source, develops in improves code. Thanks to the authors of illustrious for releasing model, thank to NoobAI team for being pioneers in open finetuning of such a scale, sharing experience, raising and solving issues that previously went unnoticed.Personal:Artists wish to remain anonymous for sharing private works; Few anonymous persons - donations, code, captions, etc., Soviet Cat - GPU sponsoring; Sv1. - llm access, captioning, code; K. - training code; Bakariso - datasets, testing, advices, insides; NeuroSenko - donations, testing, code; LOL2024 - a lot of unique datasets; T.,[] - datasets, testing, advises; rred, dga, Fi., ello - donations; TekeshiX - datasets. And other fellow brothers that helped. Love you so much \u2764\ufe0f.And off course everyone who made feedback and requests, it's really valuable.If I forgot to mention anyone, please notify.DonationsIf you want to support - share my models, leave feedback, make a cute picture with kemonomimi-girl. And of course, support original artists.AI is my hobby, I'm spending money on it and not begging for donations. However, it has turned into a large-scale and expensive undertaking. Consider to support to accelerate new training and researches.(Just keep in mind that I can waste it on alcohol or cosplay girls)BTC: bc1qwv83ggq8rvv07uk6dv4njs0j3yygj3aax4wg6cETH/USDT(e): 0x04C8a749F49aE8a56CB84cF0C99CD9E92eDB17dbXMR: 47F7JAyKP8tMBtzwxpoZsUVB8wzg2VrbtDKBice9FAS1FikbHEXXPof4PAb42CQ5ch8p8Hs4RvJuzPHDtaVSdQzD6ZbA5TZif you can offer gpu-time (a100+) - PM.",
+    "sd version": "Other"
+}

Stable-diffusion/rouwei_06Epsilon.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b96cd67807dd06f42315905bf4e50d5b9a1ec9d6feca1adb40f3a35fe8ba5f00
+size 6938040424

Stable-diffusion/rouwei_06Vpred.html

@@ -0,0 +1,250 @@
+
+                <head>
+                    
+                <meta charset="UTF-8">
+                    <link rel="stylesheet" type="text/css" href="/content/A1111/extensions/sd-civitai-browser-plus/style_html.css">
+                </head>
+                <div class="model-block">
+                    <h2><a href=https://civitai.com/models/950531 target="_blank" id="model_header">RouWei</a></h2>
+                    <h3 class="model-uploader">Uploaded by <a href="https://civitai.com/user/Minthybasis" target="_blank">Minthybasis</a><div class="avatar"><img src=https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/c3116f85-2b3b-4daf-a138-aed342f0cf2d/width=96/Minthybasis.jpeg></div></h3>
+                    <div class="civitai-version-info" style="display:flex; flex-wrap:wrap; justify-content:space-between;">
+                        <dl id="info_block">
+                            <dt>Version</dt>
+                            <dd>0.6 vpred</dd>
+                            <dt>Base Model</dt>
+                            <dd>Illustrious</dd>
+                            <dt>Published</dt>
+                            <dd>2024-11-15</dd>
+                            <dt>Availability</dt>
+                            <dd>Public</dd>
+                            <dt>CivitAI Tags</dt>
+                            <dd>
+                                <div class="civitai-tags-container">
+                                    <span class="civitai-tag">anime</span><span class="civitai-tag">base model</span>
+                                </div>
+                            </dd>
+                            <dt>Download Link</dt>
+                            <dd><a href=https://civitai.com/api/download/models/1064242 target="_blank">https://civitai.com/api/download/models/1064242</a></dd>
+                        </dl>
+                        <div style="align-self:center; min-width:320px;">
+                            <div>
+                                <p style="line-height: 2; font-weight: bold;"><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Use the model without crediting the creator<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Sell images they generate<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Run on services that generate images for money<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Run on Civitai<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Share merges using this model<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="red"><path d="M18 6l-12 12"></path><path d="M6 6l12 12"></path></svg></span> Sell this model or merges using this model<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="red"><path d="M18 6l-12 12"></path><path d="M6 6l12 12"></path></svg></span> Have different permissions when sharing merges</p>
+                            </div>
+                        </div>
+                    </div>
+                    <input type="checkbox" id="preview-civitai-description" class="description-toggle-checkbox">
+                    <div class="model-description">
+                        <h2>Description</h2>
+                        <h2 id="in-depth-retraining-of-illustrious-to-achieve-best-prompt-adherence-knowledge-and-state-of-the-art-performance.-66mudratg">In depth retraining of Illustrious to achieve best prompt adherence, knowledge and state of the art performance.</h2><p><span style="color:rgb(134, 142, 150)">Big dreams come true</span></p><p><span style="color:rgb(134, 142, 150)">The version number is just an index of current final release, not a fraction of the planned training.</span></p><p><a target="_blank" rel="ugc" href="https://huggingface.co/Minthy/RouWei-0.8"><span style="color:rgb(134, 142, 150)">HF repo</span></a></p><p></p><h3 id="large-scale-finetune-using-gpu-cluster-with-a-dataset-of-~13m-pictures-(~4m-with-natural-text-captions)-pqrwpp4f0">Large scale finetune using gpu cluster with a dataset of ~13M pictures (~4M with natural text captions)</h3><ul><li><p>Fresh and wast knowledge about characters, concepts, styles, cultural and related things</p></li><li><p>The best prompt adherence among SDXL anime models at the moment of release</p></li><li><p>Solved main problems with tags bleeding and biases, common for Illustrious, NoobAi and other checkpoints</p></li><li><p>Excellent aesthetics and knowledge across a wide range of styles (<a target="_blank" rel="ugc" href="https://files.catbox.moe/anleso.txt">over 50,000 artists</a> (<a rel="ugc" href="https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg">examples</a>), including hundreds of unique cherry-picked datasets from private galleries, including those received from the artists themselves)</p></li><li><p>High flexibility and variety without stability tradeoff</p></li><li><p>No more annoying watermarks for popular styles thanks to clean dataset</p></li><li><p>Vibrant colors and smooth gradients without trace of burning, full range even with epsilon</p></li><li><p>Pure training from Illustrious v0.1 without involving third-party checkpoints, Loras, tweakers, etc.</p></li></ul><p></p><h3 id="there-are-also-some-issues-and-changes-compared-to-the-previous-version-please-rtfm.-aaa28dbe6">There are also some issues and changes compared to the previous version, please RTFM.</h3><p></p><p>Dataset cut-off - end of April 2025.</p><p></p><h1 id="features-and-prompting:-j89sanl8e">Features and prompting:</h1><p></p><h2 id="important-change:-f6w18r0i0">Important change:</h2><p>When you are prompting artist styles, especially mixing several, <strong><u>their tags MUST BE in a separate CLIP chunk</u></strong>. Just add BREAK after it (for A1111 and derivatives), use conditioning concat node (for Comfy) <strong><u>or at least put them in the very end</u></strong>. Otherwise, significant degradation of results is likely.</p><p></p><h2 id="basic:-aroav4616">Basic:</h2><p></p><p>The checkpoint works both with short-simple and long-complex prompts. However, if there are contradictory or weird things - unlike with others they won't be ignored affecting the output. No guide-rails, no safeguards, no lobotomy.</p><p>Just prompt what you want to see and don't prompt what shouldn't be on the picture. If you want to have a view from above - don't put ceiling into positive, if you want to have crop view with head out of frame - don't make detailed description of character facial features, and so on. Pretty simple but sometimes people are missing it.</p><p></p><p>Version 0.8 comes with advanced understanding of natural text prompts. <u>It doesn't mean that you are obligated to use it</u>, tags only - completely fine, especially because understanding of tags combinations is also improved.</p><p>Do not expect it to perform like Flux or other models based on T5 or LLM text encoders. The whole size ot SDXL checkpoint is less then only that text encoder, in addition illustrious-v0.1 which is used as the base completely forgot a lot of general things from vanilla sdxl-base.</p><p>However, even in current state it works much better, allows to do new things usually impossible without external guidance, as well making manual editing, inpainting, etc more convenient.</p><p></p><p>To achieve best performance you should keep track of CLIP chunks. In SDXL the prompt is separated into a chunks of 75 (77 including BOS and EOS) tokens, that are processing by CLIP separately, and only then are concatinating and comes as conditions to unet.</p><p>If you want to specify some features for character/object and separate them from other prompt parts - make sure they are in the same chunk and optionally separate it with <code style="text-wrap: wrap">BREAK</code>. It will not solve problem of traits mixing completely, but can reduce it improving overall understanding, since text encoders on RouWei are able to process the whole sequence, not individual concepts better then others.</p><p>Dataset contains only booru-style tags and natural text expressions. Despite having a share of furries, real life photos, western media, etc. all captions have been converted to classic booru style to avoid a number of problems from mixing of different systems. So e621 tags won't be understanded properly.</p><p></p><h3 id="sampling-parameters:-7andwk13p">Sampling parameters:</h3><ul><li><p>~1 megapixel for txt2img, any AR with resolution multiple of 32 (1024x1024, 1056x, 1152x, 1216x832,...). Euler_a, 20..28steps.</p></li><li><p>CFG: for epsilon version 4..9 (7 is best), for vpred version, 3..5</p></li><li><p>Sigmas multiply may improve results a bit, CFG++ samplers work fine. LCM/PCM/DMD/... and exotic samplers untested.</p></li><li><p><strong>Some schedulers doesn't work well.</strong></p></li><li><p>Highresfix - x1.5 latent + denoise 0.6 or any gan + denoise 0.3..0.55.</p></li><li><p>For vpred version lower CFG 3..5 is needed!</p></li></ul><p></p><h3 id="for-vpred-version-lower-cfg-3..5-is-needed!-uwz9s21js"><strong>For vpred version lower CFG 3..5 is needed!</strong></h3><p></p><h2 id="quality-classification:-kw87tas72">Quality classification:</h2><p>Only 4 quality tags:</p><pre><code style="text-wrap: wrap">masterpiece, best quality</code></pre><p>for positive and</p><pre><code style="text-wrap: wrap">low quality, worst quality</code></pre><p>for negative.</p><p>Nothing else. Actually you can even omit positive and reduce negative to <code style="text-wrap: wrap">low quality</code> only, since they can affect basic style and composition.</p><p>Meta tags like <code style="text-wrap: wrap">lowres</code> have been removed and don't work, better not to use them. Low resolution images have been either removed or upscaled and cleaned with DAT depending on their importance.</p><p></p><h2 id="negative-prompt:-79eot1hhx">Negative prompt:</h2><pre><code style="text-wrap: wrap">worst quality, low quality, watermark</code></pre><p>That's all, no need of "rusty trombone", "farting on prey" and others. Do not put tags like greyscale, monochrome in negative unless you understand what are you doing. Extra tags for brightness/colors/contrast section below can be used</p><p></p><h2 id="artist-styles:-6yumq5hap">Artist styles:</h2><p></p><p><a target="_blank" rel="ugc" href="https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg">Grids with examples</a>, <a target="_blank" rel="ugc" href="https://files.catbox.moe/anleso.txt">list/wildcard</a> (also can be found in "training data").</p><p></p><h3 id="used-with-&quot;by-&quot;-it's-mandatory.-it-will-not-work-properly-without-it.-innc8cnap"><u>Used with "by " it's mandatory</u>. It will not work properly without it.</h3><p>"by " is a meta-token for styles to avoid mixing/misinterpret with tags/characters of similar or close name. This allows to have a better results for styles and at the same time avoid random style fluctuation that you may observe in other checkpoints.</p><p></p><p>Multiple give very interesting results, can be controlled with prompt weights and spells.</p><p></p><h3 id="you-must-add-break-after-artistsstyle-tags-(for-a1111)-or-concat-conditioning-(for-comfy)-or-put-them-in-the-very-end-of-your-prompt.-36i67w9c0">YOU MUST ADD <code style="text-wrap: wrap">BREAK</code> after artists/style tags (for A1111) or concat conditioning (for Comfy) or put them in the very end of your prompt.</h3><p>For example:</p><pre><code style="text-wrap: wrap">by kantoku, by wlop, best quality, masterpiece BREAK 1girl, ...</code></pre><p></p><h2 id="general-styles:-5ig9sgqza">General styles:</h2><pre><code style="text-wrap: wrap">2.5d, anime screencap, bold line, sketch, cgi, digital painting, flat colors, smooth shading, minimalistic, ink style, oil style, pastel style</code></pre><h3 id="booru-tags-styles:-uq05a7u5y">Booru tags styles:</h3><pre><code style="text-wrap: wrap">1950s (style), 1960s (style), 1970s (style), 1980s (style), 1990s (style), 2000s (style), animification, art nouveau, pinup (style), toon (style), western comics (style), nihonga, shikishi, minimalism, fine art parody</code></pre><p>and everything from <a target="_blank" rel="ugc" href="https://danbooru.donmai.us/wiki_pages/traditional_media">this group</a>.</p><p>Can be used in combinations (with artists too), with weights, both in positive and negative prompts.</p><p></p><h2 id="characters:-8htyrng9m">Characters:</h2><p>Use full name booru tag and proper formatting, like <code style="text-wrap: wrap">karin_(blue_archive)</code> -&gt; <code style="text-wrap: wrap">karin \(blue archive\)</code>, use skin tags for better reproducing, like <code style="text-wrap: wrap">karin \(bunny\) \(blue archive\)</code>. Autocomplete extension might be very useful.</p><p>Most characters are recognized just by their booru tag, but it will be more accurate if you describe their basic traits. Here you can easily redress your waifu/husbendo just by the prompt without suffering from the typical leaks of basic features.</p><p></p><h2 id="natural-text:-cq8whprlj">Natural text:</h2><p>Use it in combination with booru tags, works great. Use only natural text after typing styles and quality tags. Use just booru tags and forget about it, it's all up to you. To get best performance keep track if CLIP 75 tokens chunks.</p><p></p><p>About 4M of images in dataset had hybrid natural-text captions, made by Claude, GPT, Gemini, <a target="_blank" rel="ugc" href="https://huggingface.co/Minthy/ToriiGate-v0.4-7B">ToriiGate</a>, then refactored, cleaned and combined with tags in different variations for augmentation.</p><p></p><p>Unlike typical captions, these contains character names which is very useful. Better to keep it clean, short and convenient description works best. Better <strong><u>not</u></strong> use long and sloppy BS like</p><pre><code style="text-wrap: wrap">A mysteriously enchanting feminine entity of indeterminate yet youthful essence, whose celestial visage radiates with the ethereal luminescence of a thousand dying stars, blessed with locks cascading like the golden rivers of ancient mythology, perhaps styled in a manner reminiscent of contemporary fashion trends though not necessarily adhering to any specific aesthetic paradigm. Her eyes, pools of unfathomable depth and hue, sparkle with the wisdom of millennia yet maintain an innocent quality that defies temporal constraints...</code></pre><p>For captioning you can use <a target="_blank" rel="ugc" href="https://huggingface.co/Minthy/ToriiGate-v0.4-7B">ToriiGate</a> in short mode.</p><p></p><p>And don't expect it to be as good as flux and others, it tries very hard and after several rolls usually you can get what you want, but it is not that stable and detailed.</p><p></p><h2 id="lots-of-tailears-related-concepts:-qsxmk9gzm">Lots of Tail/Ears-related concepts:</h2><p>Oh yeah</p><pre><code style="text-wrap: wrap">tail censor, holding own tail, hugging own tail, holding another's tail, tail grab, tail raised, tail down, ears down, hand on own ear, tail around own leg, tail around penis, tailjob, tail through clothes, tail under clothes, lifted by tail, tail biting, tail penetration (including a specific indication of vaginal/anal), tail masturbation, holding with tail, panties on tail, bra on tail, tail focus, presenting own tail...</code></pre><p>(booru meaning, not e621) and many others with natural text. The majority works perfectly, some requires a lot of rolling.</p><p></p><h1 id="brightnesscolorscontrast:-lk2g7nwtj">Brightness/colors/contrast:</h1><p>You can use extra meta tags to control it:</p><pre><code style="text-wrap: wrap">low brightness, high brightness, low saturation, high saturation, low gamma, high gamma, sharp colors, soft colors, hdr, sdr</code></pre><p><a target="_blank" rel="ugc" href="https://files.catbox.moe/r8wrb6.jpg">Example</a></p><p>They work both in epsilon and vpred version and works really good.</p><p>Epsilon version relies on them too much. Without low brightness or low gamma or limited range (in negative) it might be difficult to achieve true 0,0,0 black, the same often true for white.</p><p>Both epsilon and vpred versions have like true zsnr, full range of colors and brightness without common flaws observed. But they behaves differently, just try it.</p><p></p><h1 id="vpred-version-9pr2w5h4d">Vpred version</h1><p></p><p>Main thing you need to know - lower your CFG from 7 down to 5 (or less). Otherwise, the use is similar with advantages.</p><p>It seems that starting from v0.7 vpred works flawlessly now. It shouldn't suffer from ignorance of tags close to the 75tokens chunk borders like nai. It is more difficult to get burned images - even on cfg7 usually it just over-saturated but with smooth gradients, which can be useful for some styles. Yes it can make anything from (0,0,0) to (255,255,255). You will find brightness meta tags described above quite useful for easier/lazy prompting, natural text expressions also work. To get the most dark image - put <code style="text-wrap: wrap">high brightness</code> into <u>negative</u> and/or use <code style="text-wrap: wrap">low brightness, low gamma</code> tags. If you don't like very bright skin on dark background and want to reduce contrast (or on the contrary, enhance the effect) - use hdr/sdr in negative/positive.</p><p>It was reported that in rare cases on some prompts there is a drop in contrast. Looks like other vpred models have same behaviour with such prompts, adding a "separator" closer to the border of the 75-token chunk fixes this. However, with 0.7 I haven't encountered this myself.</p><p></p><p><span style="color:rgb(134, 142, 150)">To launch vpred version you will need dev build of A1111, Comfy (with special loader node), Forge or Reforge. Just use same parameters (Euler a, cfg 3..5, 20..28 steps) like epsilon. No need to use Cfg rescale, but you can try it, cfg++ works great.</span></p><p></p><h1 id="base-model:-lwa0dfdbf">Base model:</h1><p>The model here has a small unet polishint after main training to improve small details, bump up resolution and others. Hovewer, you may be also interested into a RouWei-Base, which sometimes can perform better at complex prompts despite having minor mistakes in small details. It also comes in FP32, for example if you want to use fp32 text encoder nodes in Comfy, merge it or finetune.</p><p>It can be found in <a target="_blank" rel="ugc" href="https://huggingface.co/Minthy/RouWei-0.8">Huggingface repo</a></p><p></p><h1 id="known-issues:-jxgt6yemu">Known issues:</h1><p>Off course there are:</p><ul><li><p>Artists and style tags must be seperated into a different chunk from main prompt or come very last</p></li><li><p>There may be some positional or combinational bias in rare cases, but it's not yet clear.</p></li><li><p>There are some complaints about few of the general styles.</p></li><li><p>Epsilon version relies too much on brightness meta tags, sometimes you will need to use them to get desired brightness shift</p></li><li><p>Some newly added styles/characters might be not as good and disctinct as they deserve to</p></li><li><p>To be discovered</p><p></p></li><li><p></p></li></ul><h3 id="requests-for-artistscharacters-in-future-models-are-open.-if-you-find-artistcharacterconcept-that-perform-weak-inaccurate-or-has-strong-watermark-please-report-will-add-them-explicitly.-follow-for-a-new-versions.-kgc39ed4e">Requests for artists/characters in future models are open. If you find artist/character/concept that perform weak, inaccurate or has strong watermark - please report, will add them explicitly. Follow for a new versions.</h3><p></p><h2 id="join-the-discord-server-9lpaojolv"><a target="_blank" rel="ugc" href="https://discord.gg/ZXHENAhqE9">JOIN THE DISCORD SERVER</a></h2><p></p><h2 id="license:-0hpy342uh">License:</h2><p>Same as illustrious. Fell free to use in your merges, finetunes, ets. but please leave a link or mention, it is mandatory</p><p></p><h2 id="how-it's-made-zc68i8qve">How it's made</h2><p>I'll consider to make a report or something like it later. For sure.</p><p>In short, 98% of work is related to dataset preparations. Instead of blindly relying on loss-weighting based on tag frequency from nai paper, a custom guided loss-weighting implementation along with asynchronous collator for balancing have been used. Ztsnr (or close to it) with Epsilon prediction was achieved using noise scheduler augmentation.</p><p></p><p>Spent compute - over 8k hours of H100 (apart from research and fail attempts)</p><p></p><h1 id="thanks:-4tsy2pj9q">Thanks:</h1><p>First of all I'd like to acknowledge everyone who supports open source, develops in improves code. Thanks to the authors of illustrious for releasing model, thank to NoobAI team for being pioneers in open finetuning of such a scale, sharing experience, raising and solving issues that previously went unnoticed.</p><h3 id="personal:-9a91vqx2i">Personal:</h3><p>Artists wish to remain anonymous for sharing private works; Few anonymous persons - donations, code, captions, etc., Soviet Cat - GPU sponsoring; Sv1. - llm access, captioning, code; K. - training code; Bakariso - datasets, testing, advices, insides; NeuroSenko - donations, testing, code; LOL2024 - a lot of unique datasets; T.,[] - datasets, testing, advises; rred, dga, Fi., ello - donations; TekeshiX - datasets. And other fellow brothers that helped. Love you so much ❤️.</p><p>And off course everyone who made feedback and requests, it's really valuable.</p><p><span style="color:rgb(134, 142, 150)">If I forgot to mention anyone, please notify.</span></p><p></p><h2 id="donations-jdo27wd0z">Donations</h2><p>If you want to support - share my models, leave feedback, make a cute picture with kemonomimi-girl. And of course, support original artists.</p><p>AI is my hobby, I'm spending money on it and not begging for donations. However, it has turned into a large-scale and expensive undertaking. Consider to support to accelerate new training and researches.</p><p><span style="color:rgb(134, 142, 150)">(Just keep in mind that I can waste it on alcohol or cosplay girls)</span></p><p>BTC: bc1qwv83ggq8rvv07uk6dv4njs0j3yygj3aax4wg6c</p><p>ETH/USDT(e): 0x04C8a749F49aE8a56CB84cF0C99CD9E92eDB17db</p><p>XMR: 47F7JAyKP8tMBtzwxpoZsUVB8wzg2VrbtDKBice9FAS1FikbHEXXPof4PAb42CQ5ch8p8Hs4RvJuzPHDtaVSdQzD6ZbA5TZ</p><p>if you can offer gpu-time (a100+) - PM.</p><p></p><p></p><p></p><p></p><p></p>
+                    </div>
+                    <label for="preview-civitai-description" class="description-toggle-label"></label>
+                </div>
+                <div align=center><div class="sampleimgs"><input type="radio" name="zoomRadio" id="resetZoom" class="zoom-radio" checked>
+                    <div class="model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_0" class="zoom-radio">
+                    <label for="zoomRadiopreview_0" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/f96c7ce9-0eaf-4461-bee3-176a008ee8d2/width=1536/40258578.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/f96c7ce9-0eaf-4461-bee3-176a008ee8d2/width=1536/40258578.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>masterpiece, best quality, good quality,
+1girl, konpaku youmu, half-closed eyes, solo, black hairband, (konpaku youmu (ghost):1.2), hairband, green vest, puffy sleeves, upper body, bow, black bowtie, black bow, short hair, vest, bowtie, ribbon, leaf, shirt, wind, short sleeves, black background, looking down, white shirt, hitodama, black ribbon, hair ribbon, puffy short sleeves, grey hair, simple background, floating hair, closed mouth, expressionless, wing collar, hitodama print, buttons, green eyes, collared shirt, falling leaves, aqua eyes, glowing eyes, pale skin, (skeleton, split screen, straight-on, android, crank, gauge, gears, glowing, light bulb, mechabare, sparkle, x-ray:1.2), 
+&lt;lora:youmuanon_noob-vpred0.1_flawless_ds_v2:1&gt;</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, bad quality, lowres, jpeg artifacts, bad anatomy, bad hands, multiple views, signature, watermark, censored, covered navel, limited range, high brightness, high contrast,</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>4147779862</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1024x1024</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred_rc1</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>5</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_0">
+                                    <label class="tab-label" for="chckpreview_0">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl-vae-fp16-fix.safetensors</dd></div><div class="civitai-meta"><dt>Ensd</dt><dd>31337</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;62c7c729ad&#x27;, &#x27;model&#x27;: &#x27;e63d568c63&#x27;, &#x27;lora:youmuanon_noob-vpred0.1_flawless_ds_v2&#x27;: &#x27;7b7e860a9bae&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.3.6dev2-v1.10.1RC-latest-1453-g0fd79513</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;7b7e860a9bae&#x27;, &#x27;name&#x27;: &#x27;youmuanon_noob-vpred0.1_flawless_ds_v2&#x27;, &#x27;type&#x27;: &#x27;lora&#x27;, &#x27;weight&#x27;: 1}, {&#x27;hash&#x27;: &#x27;e63d568c63&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred_rc1&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>e63d568c63</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Align Your Steps</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>2x-AnimeSharpV3</dd></div><div class="civitai-meta"><dt>Adetailer model</dt><dd>face_yolov8n.pt</dd></div><div class="civitai-meta"><dt>Adetailer version</dt><dd>24.9.0</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Sigmas_merge_mode</dt><dd>Multiply</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.5</dd></div><div class="civitai-meta"><dt>Sigmas_mult_factor</dt><dd>2</dd></div><div class="civitai-meta"><dt>Adetailer mask blur</dt><dd>8</dd></div><div class="civitai-meta"><dt>Adetailer confidence</dt><dd>0.3</dd></div><div class="civitai-meta"><dt>Sigmas_merge_enabled</dt><dd>True</dd></div><div class="civitai-meta"><dt>Adetailer dilate erode</dt><dd>4</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Adetailer inpaint padding</dt><dd>32</dd></div><div class="civitai-meta"><dt>Adetailer negative prompt</dt><dd>{}</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div><div class="civitai-meta"><dt>Adetailer denoising strength</dt><dd>0.4</dd></div><div class="civitai-meta"><dt>Old prompt editing timelines</dt><dd>True</dd></div><div class="civitai-meta"><dt>Adetailer inpaint only masked</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_1" class="zoom-radio">
+                    <label for="zoomRadiopreview_1" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/76198546-76cd-4bf4-9fc8-f935da633bbf/width=1328/40258585.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/76198546-76cd-4bf4-9fc8-f935da633bbf/width=1328/40258585.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>(by deel \(rkeg\):0.8), best quality, masterpiece,
+1girl, folinic \(arknights\), cat girl, brown hair, multicolored hair, cat ears, surprised, blush, hair flower, holding papers, BREAK
+indoors, scientific lab, glass floor, intricate interior, computers, research facility</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, watermark</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>2290774044</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>832x1216</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_1">
+                                    <label class="tab-label" for="chckpreview_1">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;84c8888446&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;84c8888446&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>84c8888446</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.6</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_2" class="zoom-radio">
+                    <label for="zoomRadiopreview_2" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/5033de43-81ec-418e-9429-f2a7425353c4/width=1728/40258592.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/5033de43-81ec-418e-9429-f2a7425353c4/width=1728/40258592.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by ratatatat74, by tarakanovich,cube85, masterpiece, best quality,
+1girl, fake animal ears, cat ears, long black hair, large breasts, erect nipples, completely nude, butt plug, black thighhighs, garter straps, squatting, reverse cowgirl position, girl on top, from behind, ass, mature female, pussy juice, ass grab, large penis, vaginal, pov hands, tally, (slap mark:0.7),
+indoors, bedroom, dark night, dimly lit room, dark blue theme, low brightness, low gamma, sdr</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, limited range</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>1178610355</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1152x896</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_2">
+                                    <label class="tab-label" for="chckpreview_2">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_3" class="zoom-radio">
+                    <label for="zoomRadiopreview_3" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/9af10cd7-4513-4147-a4f2-033f46afd3e0/width=1824/40258577.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/9af10cd7-4513-4147-a4f2-033f46afd3e0/width=1824/40258577.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>(by ratatatat74:0.7), by tarakanovich, cube85, masterpiece, best quality,
+1girl, vulpisfoglia \(arknights\), fox girl, fox ears, large ears, ears down, blonde hair, large breasts, yellow eyes, fox tail, nightgown, see-through, bottomless, from behind, tail censor, (scar on animal ear:0.8), white panties, naughty face,
+all fours, fox shadow puppet, glowing eyes,
+black background, low brightness, low gamma</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, extra ears</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>1556719420</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1216x832</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>20</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_3">
+                                    <label class="tab-label" for="chckpreview_3">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_4" class="zoom-radio">
+                    <label for="zoomRadiopreview_4" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/0d48a8dd-f3d3-42f6-b79d-de7de7b52139/width=1200/40258590.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/0d48a8dd-f3d3-42f6-b79d-de7de7b52139/width=1200/40258590.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>masterpiece, best quality, by merrytail, by sub-res, by kie \(wylee2212\),
+1girl, fox devil \(chainsaw man\), fox girl, fox ears, fox tail, whisker markings, long white hair, red eyes, (extra eyes:1.3), fangs, evil smile, white kimono with floral print, collarbone, hair bun, fox shadow puppet, cowboy shot, posing,
+black background, low brightness</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, limited range, monster, 1other, animal, fox</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>1376750140</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>800x1344</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_4">
+                                    <label class="tab-label" for="chckpreview_4">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_5" class="zoom-radio">
+                    <label for="zoomRadiopreview_5" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/209ff688-e01d-41d4-b842-a3ffadd3741f/width=1824/40258575.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/209ff688-e01d-41d4-b842-a3ffadd3741f/width=1824/40258575.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>masterpiece, best quality, by done \(donezumi\),
+1girl, kon \(tokyo ravens\), tokyo ravens, fox girl, fox ears, fox tail, white hair, thick eyebrows, pale skin, blue eyes, white aqua japanese clothes, detached sleeves, ribbon trim, fox shadow puppet, pom pom \(clothes\), black shorts, white thighhighs, hips, happy, BREAK floating in air, legs apart, posing, full body, wide shot,
+white background, high brightness</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>2168859150</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1216x832</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>20</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_5">
+                                    <label class="tab-label" for="chckpreview_5">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;84c8888446&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;84c8888446&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>84c8888446</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_6" class="zoom-radio">
+                    <label for="zoomRadiopreview_6" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/90435b3a-ef7d-4be2-b700-00601088c553/width=1248/40258594.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/90435b3a-ef7d-4be2-b700-00601088c553/width=1248/40258594.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by tandohark, (by personal ami:0.6), best quality, masterpiece,
+2girls, asymmetrical docking BREAK
+izuna \(blue archive\), fox girl, fox ears, fox mask, fox tail, halo, topless, denim shorts, striped bikini bra, medium breasts, sideboob, visors cap BREAK
+wakamo \(blue archive\), fox girl, fox ears, fox tail, long black hair, red halo, hair ornament, large breasts, fox mask, white bikini swimsuit, collarbone, :3 BREAK
+outdoors, beach, seashore, palms, clear blue sky</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, watermark</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>1114850771</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>832x1216</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>5</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_6">
+                                    <label class="tab-label" for="chckpreview_6">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;84c8888446&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;84c8888446&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>84c8888446</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_7" class="zoom-radio">
+                    <label for="zoomRadiopreview_7" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/23c627f5-451b-47de-a1d3-7dce53adb8d9/width=1248/40258589.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/23c627f5-451b-47de-a1d3-7dce53adb8d9/width=1248/40258589.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by akai sashimi, by nat the lich, by foxykuro, by nekoya \(liu\), masterpiece,  best quality,
+1girl, kal&#x27;tsit \(arknights\), cat girl, lynx ears, silver hair, green eyes/, naked towel, covered nipples, off shoulder, oripathy lesion \(arknights\), collarbone, shaded face, (evil smile:1.1), smug, open mouth, angry, fang, tilted head, wet,
+holding handcuffs,
+cowboy shot, motion lines,
+indoors, office, dimmed, dark background</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, watermark, dutch angle</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>2542436241</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>832x1216</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_7">
+                                    <label class="tab-label" for="chckpreview_7">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Ensd</dt><dd>31337</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_8" class="zoom-radio">
+                    <label for="zoomRadiopreview_8" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/fd31b937-db26-404f-ba61-c86eecedb37c/width=1328/40258587.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/fd31b937-db26-404f-ba61-c86eecedb37c/width=1328/40258587.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by akai sashimi, by nat the lich, by foxykuro, by nekoya \(liu\), masterpiece, best quality,
+1girl, kal&#x27;tsit \(arknights\), cat girl, lynx ears, silver hair, green eyes, short tail, completely nude, oripathy lesion \(arknights\), scars on spine, reverse cowgirl position, motion lines, from below, open mouth, smug, heavenly ass, v,
+1boy, vaginal, sex, pov, deep penetration, pussy juice trail,
+indoors, bedroom, dimmed, dark background</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, watermark, abstract background</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>3956367149</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>832x1216</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>20</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_8">
+                                    <label class="tab-label" for="chckpreview_8">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.6</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_9" class="zoom-radio">
+                    <label for="zoomRadiopreview_9" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/11c49fc0-f6fa-40b2-bb86-30e3a2c9cd0a/width=2064/40258599.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/11c49fc0-f6fa-40b2-bb86-30e3a2c9cd0a/width=2064/40258599.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by kate fox, masterpiece, best quality,
+1girl, yuel \(granblue fantasy\), fox girl, yellow eyes, fox ears, large fox tail, bikini bra, fur trim, elbow gloves, :3, wide shot, BREAK
+outdoors, torii, asian nature, epic landscape of mountains, above clouds</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, bad hands, watermark
+worst quality, low quality</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>2135703813</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1376x720</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>20</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_9">
+                                    <label class="tab-label" for="chckpreview_9">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;84c8888446&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;84c8888446&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>84c8888446</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Rescale_cfg_enabled</dt><dd>True</dd></div><div class="civitai-meta"><dt>Rescale_cfg_multiplier</dt><dd>0.7</dd></div></div></div></div></dl></div></div></div></div>
+                

Stable-diffusion/rouwei_06Vpred.json

@@ -0,0 +1,8 @@
+{
+    "sha256": "84C8888446CA93B6B8C204BB558A9F16138138A1CD0F4FDC264C18FE4C006163",
+    "modelId": 950531,
+    "modelVersionId": 1064242,
+    "activation text": "",
+    "description": "In depth retraining of Illustrious to achieve best prompt adherence, knowledge and state of the art performance.Big dreams come trueThe version number is just an index of current final release, not a fraction of the planned training.HF repo https://huggingface.co/Minthy/RouWei-0.8Large scale finetune using gpu cluster with a dataset of ~13M pictures (~4M with natural text captions)Fresh and wast knowledge about characters, concepts, styles, cultural and related thingsThe best prompt adherence among SDXL anime models at the moment of releaseSolved main problems with tags bleeding and biases, common for Illustrious, NoobAi and other checkpointsExcellent aesthetics and knowledge across a wide range of styles (over 50,000 artists https://files.catbox.moe/anleso.txt (examples https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg), including hundreds of unique cherry-picked datasets from private galleries, including those received from the artists themselves)High flexibility and variety without stability tradeoffNo more annoying watermarks for popular styles thanks to clean datasetVibrant colors and smooth gradients without trace of burning, full range even with epsilonPure training from Illustrious v0.1 without involving third-party checkpoints, Loras, tweakers, etc.There are also some issues and changes compared to the previous version, please RTFM.Dataset cut-off - end of April 2025.Features and prompting:Important change:When you are prompting artist styles, especially mixing several, their tags MUST BE in a separate CLIP chunk. Just add BREAK after it (for A1111 and derivatives), use conditioning concat node (for Comfy) or at least put them in the very end. Otherwise, significant degradation of results is likely.Basic:The checkpoint works both with short-simple and long-complex prompts. However, if there are contradictory or weird things - unlike with others they won't be ignored affecting the output. No guide-rails, no safeguards, no lobotomy.Just prompt what you want to see and don't prompt what shouldn't be on the picture. If you want to have a view from above - don't put ceiling into positive, if you want to have crop view with head out of frame - don't make detailed description of character facial features, and so on. Pretty simple but sometimes people are missing it.Version 0.8 comes with advanced understanding of natural text prompts. It doesn't mean that you are obligated to use it, tags only - completely fine, especially because understanding of tags combinations is also improved.Do not expect it to perform like Flux or other models based on T5 or LLM text encoders. The whole size ot SDXL checkpoint is less then only that text encoder, in addition illustrious-v0.1 which is used as the base completely forgot a lot of general things from vanilla sdxl-base.However, even in current state it works much better, allows to do new things usually impossible without external guidance, as well making manual editing, inpainting, etc more convenient.To achieve best performance you should keep track of CLIP chunks. In SDXL the prompt is separated into a chunks of 75 (77 including BOS and EOS) tokens, that are processing by CLIP separately, and only then are concatinating and comes as conditions to unet.If you want to specify some features for character/object and separate them from other prompt parts - make sure they are in the same chunk and optionally separate it with BREAK. It will not solve problem of traits mixing completely, but can reduce it improving overall understanding, since text encoders on RouWei are able to process the whole sequence, not individual concepts better then others.Dataset contains only booru-style tags and natural text expressions. Despite having a share of furries, real life photos, western media, etc. all captions have been converted to classic booru style to avoid a number of problems from mixing of different systems. So e621 tags won't be understanded properly.Sampling parameters:~1 megapixel for txt2img, any AR with resolution multiple of 32 (1024x1024, 1056x, 1152x, 1216x832,...). Euler_a, 20..28steps.CFG: for epsilon version 4..9 (7 is best), for vpred version, 3..5Sigmas multiply may improve results a bit, CFG++ samplers work fine. LCM/PCM/DMD/... and exotic samplers untested.Some schedulers doesn't work well.Highresfix - x1.5 latent + denoise 0.6 or any gan + denoise 0.3..0.55.For vpred version lower CFG 3..5 is needed!For vpred version lower CFG 3..5 is needed!Quality classification:Only 4 quality tags:masterpiece, best qualityfor positive andlow quality, worst qualityfor negative.Nothing else. Actually you can even omit positive and reduce negative to low quality only, since they can affect basic style and composition.Meta tags like lowres have been removed and don't work, better not to use them. Low resolution images have been either removed or upscaled and cleaned with DAT depending on their importance.Negative prompt:worst quality, low quality, watermarkThat's all, no need of \"rusty trombone\", \"farting on prey\" and others. Do not put tags like greyscale, monochrome in negative unless you understand what are you doing. Extra tags for brightness/colors/contrast section below can be usedArtist styles:Grids with examples https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg, list/wildcard https://files.catbox.moe/anleso.txt (also can be found in \"training data\").Used with \"by \" it's mandatory. It will not work properly without it.\"by \" is a meta-token for styles to avoid mixing/misinterpret with tags/characters of similar or close name. This allows to have a better results for styles and at the same time avoid random style fluctuation that you may observe in other checkpoints.Multiple give very interesting results, can be controlled with prompt weights and spells.YOU MUST ADD BREAK after artists/style tags (for A1111) or concat conditioning (for Comfy) or put them in the very end of your prompt.For example:by kantoku, by wlop, best quality, masterpiece BREAK 1girl, ...General styles:2.5d, anime screencap, bold line, sketch, cgi, digital painting, flat colors, smooth shading, minimalistic, ink style, oil style, pastel styleBooru tags styles:1950s (style), 1960s (style), 1970s (style), 1980s (style), 1990s (style), 2000s (style), animification, art nouveau, pinup (style), toon (style), western comics (style), nihonga, shikishi, minimalism, fine art parodyand everything from this group https://danbooru.donmai.us/wiki_pages/traditional_media.Can be used in combinations (with artists too), with weights, both in positive and negative prompts.Characters:Use full name booru tag and proper formatting, like karin_(blue_archive) -> karin \\(blue archive\\), use skin tags for better reproducing, like karin \\(bunny\\) \\(blue archive\\). Autocomplete extension might be very useful.Most characters are recognized just by their booru tag, but it will be more accurate if you describe their basic traits. Here you can easily redress your waifu/husbendo just by the prompt without suffering from the typical leaks of basic features.Natural text:Use it in combination with booru tags, works great. Use only natural text after typing styles and quality tags. Use just booru tags and forget about it, it's all up to you. To get best performance keep track if CLIP 75 tokens chunks.About 4M of images in dataset had hybrid natural-text captions, made by Claude, GPT, Gemini, ToriiGate https://huggingface.co/Minthy/ToriiGate-v0.4-7B, then refactored, cleaned and combined with tags in different variations for augmentation.Unlike typical captions, these contains character names which is very useful. Better to keep it clean, short and convenient description works best. Better not use long and sloppy BS likeA mysteriously enchanting feminine entity of indeterminate yet youthful essence, whose celestial visage radiates with the ethereal luminescence of a thousand dying stars, blessed with locks cascading like the golden rivers of ancient mythology, perhaps styled in a manner reminiscent of contemporary fashion trends though not necessarily adhering to any specific aesthetic paradigm. Her eyes, pools of unfathomable depth and hue, sparkle with the wisdom of millennia yet maintain an innocent quality that defies temporal constraints...For captioning you can use ToriiGate https://huggingface.co/Minthy/ToriiGate-v0.4-7B in short mode.And don't expect it to be as good as flux and others, it tries very hard and after several rolls usually you can get what you want, but it is not that stable and detailed.Lots of Tail/Ears-related concepts:Oh yeahtail censor, holding own tail, hugging own tail, holding another's tail, tail grab, tail raised, tail down, ears down, hand on own ear, tail around own leg, tail around penis, tailjob, tail through clothes, tail under clothes, lifted by tail, tail biting, tail penetration (including a specific indication of vaginal/anal), tail masturbation, holding with tail, panties on tail, bra on tail, tail focus, presenting own tail...(booru meaning, not e621) and many others with natural text. The majority works perfectly, some requires a lot of rolling.Brightness/colors/contrast:You can use extra meta tags to control it:low brightness, high brightness, low saturation, high saturation, low gamma, high gamma, sharp colors, soft colors, hdr, sdrExampleThey work both in epsilon and vpred version and works really good.Epsilon version relies on them too much. Without low brightness or low gamma or limited range (in negative) it might be difficult to achieve true 0,0,0 black, the same often true for white.Both epsilon and vpred versions have like true zsnr, full range of colors and brightness without common flaws observed. But they behaves differently, just try it.Vpred versionMain thing you need to know - lower your CFG from 7 down to 5 (or less). Otherwise, the use is similar with advantages.It seems that starting from v0.7 vpred works flawlessly now. It shouldn't suffer from ignorance of tags close to the 75tokens chunk borders like nai. It is more difficult to get burned images - even on cfg7 usually it just over-saturated but with smooth gradients, which can be useful for some styles. Yes it can make anything from (0,0,0) to (255,255,255). You will find brightness meta tags described above quite useful for easier/lazy prompting, natural text expressions also work. To get the most dark image - put high brightness into negative and/or use low brightness, low gamma tags. If you don't like very bright skin on dark background and want to reduce contrast (or on the contrary, enhance the effect) - use hdr/sdr in negative/positive.It was reported that in rare cases on some prompts there is a drop in contrast. Looks like other vpred models have same behaviour with such prompts, adding a \"separator\" closer to the border of the 75-token chunk fixes this. However, with 0.7 I haven't encountered this myself.To launch vpred version you will need dev build of A1111, Comfy (with special loader node), Forge or Reforge. Just use same parameters (Euler a, cfg 3..5, 20..28 steps) like epsilon. No need to use Cfg rescale, but you can try it, cfg++ works great.Base model:The model here has a small unet polishint after main training to improve small details, bump up resolution and others. Hovewer, you may be also interested into a RouWei-Base, which sometimes can perform better at complex prompts despite having minor mistakes in small details. It also comes in FP32, for example if you want to use fp32 text encoder nodes in Comfy, merge it or finetune.It can be found in Huggingface repo https://huggingface.co/Minthy/RouWei-0.8Known issues:Off course there are:Artists and style tags must be seperated into a different chunk from main prompt or come very lastThere may be some positional or combinational bias in rare cases, but it's not yet clear.There are some complaints about few of the general styles.Epsilon version relies too much on brightness meta tags, sometimes you will need to use them to get desired brightness shiftSome newly added styles/characters might be not as good and disctinct as they deserve toTo be discoveredRequests for artists/characters in future models are open. If you find artist/character/concept that perform weak, inaccurate or has strong watermark - please report, will add them explicitly. Follow for a new versions.JOIN THE DISCORD SERVER https://discord.gg/ZXHENAhqE9License:Same as illustrious. Fell free to use in your merges, finetunes, ets. but please leave a link or mention, it is mandatoryHow it's madeI'll consider to make a report or something like it later. For sure.In short, 98% of work is related to dataset preparations. Instead of blindly relying on loss-weighting based on tag frequency from nai paper, a custom guided loss-weighting implementation along with asynchronous collator for balancing have been used. Ztsnr (or close to it) with Epsilon prediction was achieved using noise scheduler augmentation.Spent compute - over 8k hours of H100 (apart from research and fail attempts)Thanks:First of all I'd like to acknowledge everyone who supports open source, develops in improves code. Thanks to the authors of illustrious for releasing model, thank to NoobAI team for being pioneers in open finetuning of such a scale, sharing experience, raising and solving issues that previously went unnoticed.Personal:Artists wish to remain anonymous for sharing private works; Few anonymous persons - donations, code, captions, etc., Soviet Cat - GPU sponsoring; Sv1. - llm access, captioning, code; K. - training code; Bakariso - datasets, testing, advices, insides; NeuroSenko - donations, testing, code; LOL2024 - a lot of unique datasets; T.,[] - datasets, testing, advises; rred, dga, Fi., ello - donations; TekeshiX - datasets. And other fellow brothers that helped. Love you so much \u2764\ufe0f.And off course everyone who made feedback and requests, it's really valuable.If I forgot to mention anyone, please notify.DonationsIf you want to support - share my models, leave feedback, make a cute picture with kemonomimi-girl. And of course, support original artists.AI is my hobby, I'm spending money on it and not begging for donations. However, it has turned into a large-scale and expensive undertaking. Consider to support to accelerate new training and researches.(Just keep in mind that I can waste it on alcohol or cosplay girls)BTC: bc1qwv83ggq8rvv07uk6dv4njs0j3yygj3aax4wg6cETH/USDT(e): 0x04C8a749F49aE8a56CB84cF0C99CD9E92eDB17dbXMR: 47F7JAyKP8tMBtzwxpoZsUVB8wzg2VrbtDKBice9FAS1FikbHEXXPof4PAb42CQ5ch8p8Hs4RvJuzPHDtaVSdQzD6ZbA5TZif you can offer gpu-time (a100+) - PM.",
+    "sd version": "Other"
+}

Stable-diffusion/rouwei_06Vpred.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:84c8888446ca93b6b8c204bb558a9f16138138a1cd0f4fdc264c18fe4c006163
+size 6938040512

Stable-diffusion/rouwei_v07Epsilon.json

@@ -3,6 +3,6 @@
     "modelId": 950531,
     "modelVersionId": 1330983,
     "activation text": "",
-    "description": "Large scale finetune of Illustrious with state of the art performance.(tl/dr: Works exactly as it should without flaws you might encounter in other checkpoints.)V0.7 comes with the dataset of 7M images (~2M with natural text captions):New knowledge for characters, concepts, stylesMost of the things that was not good enough in 0.6 now on decent levelEven better prompting with good segmentation and natural phrases understanding (now it's even possible to easily make one character to cosplay other without typical issues 1, 2, 3)Better stability and broader coverage, less biasesExcelent performance for both very simple and complex promptsLots of general changes and fixesv0.7 vpred will be released soonKey advantages:Easy and convenient promptingGreat aesthetic, anatomy, stability along with versatilityVibrant colors and smooth gradients without trace of burningFull brightness range even with epsilon35k+ artist styles, many general styles, almost any characterPure training without any merged weights, loras, tweakers, etc., you can add your own if neededAn addition to mentioned, comparing with vanilla Illustrious and NoobAI:No more annoying watermarksNo tags bleed and better segmentation (1, 2, 3 (nsfw))No characters bleed and related side effects (unwanted outfits, style, composition changes)No spawning of strange creatures, sfx on background or extra pair of breastsBetter coherence, prompt following, anatomy (significant boost over illustrious, slight over noob)Artist styles look exactly as they should (and lots of new added)Better prompt following without ignoring tags and need of (higher weights:1.4)Forget about long scizo-negativeStable style without random fluctuations on different seedsNew charactersDataset cut-off - 20th December 2024.Features and prompting:Basic:The checkpoint works both with short-simple and long-complex prompts. However, if there are contradictory or weird things - unlike with other checkpoints, they won't be ignored affecting the output. No guide-rails, no safeguards, no lobotomy.Just prompt what you want to see and don't prompt what shouldn't be on the picture. If you want to have a view from above - don't put ceiling into positive, if you want to have crop view with head out of frame - don't make detailed description of character facial features, and so on. Pretty simple but sometimes people are missing it.Version 0.7 comes with several improvements in prompt understanding and segmentation.However, because SDXL is based on CLIP text encoders which have limit of 75 (77) tokens for input, it is quite limited due to prompt chunking. If you want to specify some features for character/object and separate them from other prompt parts - make sure they are in the same chunk and optionally separate it with BREAK.It will not solve problem of traits mixing completely, but can reduce it improving overall understanding, since text encoders on RouWei are able to process the whole sequence, not individual concepts better then others.Dataset contains only booru-style tags and (simplified) natural text expressions. Despite having a share of furries, all captions have been converted to classic booru style to avoid a number of problems that may arise when mixing different systems. So e621 tags won't be understanded properly.Sampling parameters:~1 megapixel for txt2img, any AR with resolution multiple of 32 (1024x1024, 1056x, 1152x, 1216x832,...). Euler_a, 20..28steps.CFG: for epsilon version 4..9 (7 is best), for vpred version, 3..5Sigmas multiply may improve results a bit, CFG++ samplers work fine. LCM/PCM/DMD/... and exotic samplers untested.Highresfix - x1.5 latent + denoise 0.6 or any gan + denoise 0.3..0.55.For vpred version lower CFG 3..5 is needed!For vpred version lower CFG 3..5 is needed!Quality classification:Only 4 quality tags:masterpiece, best qualitylow quality, worst qualityNothing else. Meta tags like lowres have been removed, better not to use them. Low resolution images have been either removed or upscaled and cleaned with DAT depending on their importance.Negative prompt:worst quality, low quality, watermarkThat's all, no need of \"rusty trombone\", \"farting on prey\" and others. Do not put tags like greyscale, monochrome in negative unless you understand what are you doing. It will lead to burning and over-saturation, colors are fine out of box.Hovewer if you want to boost or adjust it - please use extra tags from brightness/colors/contrast section below.Artist styles:Grids with examples https://mega.nz/folder/lGQSVDyZ#uzuzfSlPRzgsWsxPQPbuhw, list https://files.catbox.moe/q6q80c.txt (also can be found in \"training data\").Used with \"by \" it's mandatory. It will not work properly without it.\"by \" is used as meta-token for styles to avoid mixing/misinterpret with tags/characters of similar or close name. This allows to have a better results for styles and at the same time avoid random style fluctuation that you may observe in some other checkpoints.Multiple give very interesting results, can be controlled with prompt weights.General styles:2.5d, anime screencap, bold line, sketch, cgi, digital painting, flat colors, smooth shading, minimalistic, ink style, oil style, pastel styleBooru tags styles:1950s (style), 1960s (style), 1970s (style), 1980s (style), 1990s (style), 1990s (style), animification, art nouveau, pinup (style), toon (style), western comics (style), nihonga, shikishi, minimalism, fine art parodyand everything from this group https://danbooru.donmai.us/wiki_pages/traditional_media.Can be used in combinations (with artists too), with weights, both in positive and negative prompts.Characters:Use full name booru tag and proper formatting, like karin_(blue_archive) -> karin \\(blue_archive\\), use skin tags for better reproducing, like karin \\(bunny \\(blue_archive\\). Autocomplete extension might be very useful.Most characters are recognized just by their booru tag, but it will be more accurate if you describe their basic traits. Here you can easily redress your waifu/husbendo just by the prompt without suffering from the typical leaks of basic features.Natural text:Use it in combination with booru tags, works great. Use only natural text after typing styles and quality tags. Use just booru tags and forget about it, it's all up to you. To get best performance keep track if CLIP 75 tokens chunks.About 2M images in dataset have hybrid natural-text captions made by Opus-Vision, GPT-4o and ToriiGate https://huggingface.co/Minthy/ToriiGate-v0.4-7B, which improves segmentation, allows to get some understanding and provides a response to chain of phrases, not single words. But overall performance is not even close to flux or sd3.5.Lots of Tail/Ears-related concepts:tail censor, holding own tail, hugging own tail, holding another's tail, tail grab, tail raised, tail down, ears down, hand on own ear, tail around own leg, tail around penis, tail through clothes, tail under clothes, lifted by tail, tail biting, tail insertion, tail masturbation, holding with tail, ...(booru meaning, not e621) and many others with natural text. The majority works perfectly, some requires rolling.Brightness/colors/contrast:You can use extra meta tags to control it:low brightness, high brightness, low saturation, high saturation, low gamma, high gamma, sharp colors, soft colors, hdr, sdrExampleThey work both in epsilon and vpred version and works really good.Unfortunately here is an issue - the model (epsilon) relies on them too much. Without low brightness or low gamma or limited range (in negative) it might be difficult to achieve true 0,0,0 black, the same often true for white.Both epsilon and vpred versions have like true zsnr, full range of colors and brightness without common flaws observed. But they behaves differently, just try it.Vpred versionVpred version of 0.7 will be released soon.For version 0.6.1:After 0.6.1 update it works much better and same issues as before are not observed. It might react on prompt differently and clip chunks composition here have a bigger influence, but same observed for all vpred checkpoints so far.At the moment of release this is probably the only vpred model that runs okay and doesn't suffer from burned colors, limited range, need of extra tweaks, rescales, adjustments and so on (default parameters: 1, 2, cfg rescale: 1, 2, 3 old examples but still may be relevant). It tends to have same like NAI3 behaviour with wrong skin colors and large fillups with red/yellow/blue if you boost CFG too much. Full experience lmao.To launch vpred version you will need dev build of A1111, comfy (with special loader node) or Reforge. Just use same parameters (Euler a, cfg 3..5, 20..28 steps) like epsilon. No need to use Cfg rescale, but you can try it.It does not rely on brightness tag too much and can make very dark-very bright images based on prompt content. But it's still very convenient to use them, especially to fix some possible artist/concept biases or to lazily achieve what you want.Base model:Epsilon and vpred versions here have a brief aesthetic polishing after main training to improve small details and coherence. If you want to use RouWei in merges, extract something without bringing that last things, or finetune it - you can find base version of RouWei on higgingface (soon).Known issues:Off course there are:Epsilon version relies too much on brightness meta tags, sometimes you will need to use them to get desired brightness shiftSome newly added styles/characters might be not as good and disctinct as they deserve toInferior in furry-related knowledge compared to NoobAiTo be discoveredRequests for artists/characters in future models are open. If you find artist/character/concept that perform weak, inaccurate or has strong watermark - please report, will add them explicitly. Follow for a new versions.JOIN THE DISCORD SERVER https://discord.gg/ZXHENAhqE9License:Same as illustrious. Fell free to use in your merges, finetunes, ets. just please leave a link.How it's madeI'll consider to make a report or something like it later. For sure.In short, 98% of work is related to dataset preparations. Instead of blindly relying on loss-weighting based on tag frequency from nai paper, a custom guided loss-weighting implementation along with asynchronous collator for balancing have been used. Ztsnr (or close to it) with Epsilon prediction was achieved using noise scheduler augmentation.Spent compute - about 25days of 4xH100 (apart from research and fail attempts)Thanks:First of all I'd like to acknowledge everyone who supports open source, develops in improves code. Thanks to the authors of illustrious for releasing model, thank to NoobAI team for being pioneers in open finetuning of such a scale, sharing experience, raising and solving issues that previously went unnoticed.Personal:Artists wish to remain anonymous for sharing private works; Few anonymous persons - donations, code, captions, etc., Soviet Cat - GPU sponsoring; Sv1. - llm access, captioning, code; K. - training code; Bakariso - datasets, testing, advices, insides; NeuroSenko - donations, testing, code; LOL2024 - a lot of unique datasets; T.,[] - datasets, testing, advises; rred, dga, Fi., ello - donations; other fellow brothers that helped. Love you so much \u2764\ufe0f.And off course everyone who made feedback and requests, it's really valuable.If I forgot to mention anyone, please notify.DonationsIf you want to support - share my models, leave feedback, make a cute picture with kemonomimi-girl. And of course, support original artists.AI is my hobby, I'm spending money on it and not begging for donations. However, it has turned into a large-scale and expensive undertaking. Consider to support to accelerate new training and researches.(Just keep in mind that I can waste it on alcohol or cosplay girls)BTC: bc1qwv83ggq8rvv07uk6dv4njs0j3yygj3aax4wg6cETH/USDT(e): 0x04C8a749F49aE8a56CB84cF0C99CD9E92eDB17dbXMR: 47F7JAyKP8tMBtzwxpoZsUVB8wzg2VrbtDKBice9FAS1FikbHEXXPof4PAb42CQ5ch8p8Hs4RvJuzPHDtaVSdQzD6ZbA5TZif you can offer gpu-time (a100+) - PM.",
+    "description": "In depth retraining of Illustrious to achieve best prompt adherence, knowledge and state of the art performance.Big dreams come trueThe version number is just an index of current final release, not a fraction of the planned training.HF repo https://huggingface.co/Minthy/RouWei-0.8Large scale finetune using gpu cluster with a dataset of ~13M pictures (~4M with natural text captions)Fresh and wast knowledge about characters, concepts, styles, cultural and related thingsThe best prompt adherence among SDXL anime models at the moment of releaseSolved main problems with tags bleeding and biases, common for Illustrious, NoobAi and other checkpointsExcellent aesthetics and knowledge across a wide range of styles (over 50,000 artists https://files.catbox.moe/anleso.txt (examples https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg), including hundreds of unique cherry-picked datasets from private galleries, including those received from the artists themselves)High flexibility and variety without stability tradeoffNo more annoying watermarks for popular styles thanks to clean datasetVibrant colors and smooth gradients without trace of burning, full range even with epsilonPure training from Illustrious v0.1 without involving third-party checkpoints, Loras, tweakers, etc.There are also some issues and changes compared to the previous version, please RTFM.Dataset cut-off - end of April 2025.Features and prompting:Important change:When you are prompting artist styles, especially mixing several, their tags MUST BE in a separate CLIP chunk. Just add BREAK after it (for A1111 and derivatives), use conditioning concat node (for Comfy) or at least put them in the very end. Otherwise, significant degradation of results is likely.Basic:The checkpoint works both with short-simple and long-complex prompts. However, if there are contradictory or weird things - unlike with others they won't be ignored affecting the output. No guide-rails, no safeguards, no lobotomy.Just prompt what you want to see and don't prompt what shouldn't be on the picture. If you want to have a view from above - don't put ceiling into positive, if you want to have crop view with head out of frame - don't make detailed description of character facial features, and so on. Pretty simple but sometimes people are missing it.Version 0.8 comes with advanced understanding of natural text prompts. It doesn't mean that you are obligated to use it, tags only - completely fine, especially because understanding of tags combinations is also improved.Do not expect it to perform like Flux or other models based on T5 or LLM text encoders. The whole size ot SDXL checkpoint is less then only that text encoder, in addition illustrious-v0.1 which is used as the base completely forgot a lot of general things from vanilla sdxl-base.However, even in current state it works much better, allows to do new things usually impossible without external guidance, as well making manual editing, inpainting, etc more convenient.To achieve best performance you should keep track of CLIP chunks. In SDXL the prompt is separated into a chunks of 75 (77 including BOS and EOS) tokens, that are processing by CLIP separately, and only then are concatinating and comes as conditions to unet.If you want to specify some features for character/object and separate them from other prompt parts - make sure they are in the same chunk and optionally separate it with BREAK. It will not solve problem of traits mixing completely, but can reduce it improving overall understanding, since text encoders on RouWei are able to process the whole sequence, not individual concepts better then others.Dataset contains only booru-style tags and natural text expressions. Despite having a share of furries, real life photos, western media, etc. all captions have been converted to classic booru style to avoid a number of problems from mixing of different systems. So e621 tags won't be understanded properly.Sampling parameters:~1 megapixel for txt2img, any AR with resolution multiple of 32 (1024x1024, 1056x, 1152x, 1216x832,...). Euler_a, 20..28steps.CFG: for epsilon version 4..9 (7 is best), for vpred version, 3..5Sigmas multiply may improve results a bit, CFG++ samplers work fine. LCM/PCM/DMD/... and exotic samplers untested.Some schedulers doesn't work well.Highresfix - x1.5 latent + denoise 0.6 or any gan + denoise 0.3..0.55.For vpred version lower CFG 3..5 is needed!For vpred version lower CFG 3..5 is needed!Quality classification:Only 4 quality tags:masterpiece, best qualityfor positive andlow quality, worst qualityfor negative.Nothing else. Actually you can even omit positive and reduce negative to low quality only, since they can affect basic style and composition.Meta tags like lowres have been removed and don't work, better not to use them. Low resolution images have been either removed or upscaled and cleaned with DAT depending on their importance.Negative prompt:worst quality, low quality, watermarkThat's all, no need of \"rusty trombone\", \"farting on prey\" and others. Do not put tags like greyscale, monochrome in negative unless you understand what are you doing. Extra tags for brightness/colors/contrast section below can be usedArtist styles:Grids with examples https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg, list/wildcard https://files.catbox.moe/anleso.txt (also can be found in \"training data\").Used with \"by \" it's mandatory. It will not work properly without it.\"by \" is a meta-token for styles to avoid mixing/misinterpret with tags/characters of similar or close name. This allows to have a better results for styles and at the same time avoid random style fluctuation that you may observe in other checkpoints.Multiple give very interesting results, can be controlled with prompt weights and spells.YOU MUST ADD BREAK after artists/style tags (for A1111) or concat conditioning (for Comfy) or put them in the very end of your prompt.For example:by kantoku, by wlop, best quality, masterpiece BREAK 1girl, ...General styles:2.5d, anime screencap, bold line, sketch, cgi, digital painting, flat colors, smooth shading, minimalistic, ink style, oil style, pastel styleBooru tags styles:1950s (style), 1960s (style), 1970s (style), 1980s (style), 1990s (style), 2000s (style), animification, art nouveau, pinup (style), toon (style), western comics (style), nihonga, shikishi, minimalism, fine art parodyand everything from this group https://danbooru.donmai.us/wiki_pages/traditional_media.Can be used in combinations (with artists too), with weights, both in positive and negative prompts.Characters:Use full name booru tag and proper formatting, like karin_(blue_archive) -> karin \\(blue archive\\), use skin tags for better reproducing, like karin \\(bunny\\) \\(blue archive\\). Autocomplete extension might be very useful.Most characters are recognized just by their booru tag, but it will be more accurate if you describe their basic traits. Here you can easily redress your waifu/husbendo just by the prompt without suffering from the typical leaks of basic features.Natural text:Use it in combination with booru tags, works great. Use only natural text after typing styles and quality tags. Use just booru tags and forget about it, it's all up to you. To get best performance keep track if CLIP 75 tokens chunks.About 4M of images in dataset had hybrid natural-text captions, made by Claude, GPT, Gemini, ToriiGate https://huggingface.co/Minthy/ToriiGate-v0.4-7B, then refactored, cleaned and combined with tags in different variations for augmentation.Unlike typical captions, these contains character names which is very useful. Better to keep it clean, short and convenient description works best. Better not use long and sloppy BS likeA mysteriously enchanting feminine entity of indeterminate yet youthful essence, whose celestial visage radiates with the ethereal luminescence of a thousand dying stars, blessed with locks cascading like the golden rivers of ancient mythology, perhaps styled in a manner reminiscent of contemporary fashion trends though not necessarily adhering to any specific aesthetic paradigm. Her eyes, pools of unfathomable depth and hue, sparkle with the wisdom of millennia yet maintain an innocent quality that defies temporal constraints...For captioning you can use ToriiGate https://huggingface.co/Minthy/ToriiGate-v0.4-7B in short mode.And don't expect it to be as good as flux and others, it tries very hard and after several rolls usually you can get what you want, but it is not that stable and detailed.Lots of Tail/Ears-related concepts:Oh yeahtail censor, holding own tail, hugging own tail, holding another's tail, tail grab, tail raised, tail down, ears down, hand on own ear, tail around own leg, tail around penis, tailjob, tail through clothes, tail under clothes, lifted by tail, tail biting, tail penetration (including a specific indication of vaginal/anal), tail masturbation, holding with tail, panties on tail, bra on tail, tail focus, presenting own tail...(booru meaning, not e621) and many others with natural text. The majority works perfectly, some requires a lot of rolling.Brightness/colors/contrast:You can use extra meta tags to control it:low brightness, high brightness, low saturation, high saturation, low gamma, high gamma, sharp colors, soft colors, hdr, sdrExampleThey work both in epsilon and vpred version and works really good.Epsilon version relies on them too much. Without low brightness or low gamma or limited range (in negative) it might be difficult to achieve true 0,0,0 black, the same often true for white.Both epsilon and vpred versions have like true zsnr, full range of colors and brightness without common flaws observed. But they behaves differently, just try it.Vpred versionMain thing you need to know - lower your CFG from 7 down to 5 (or less). Otherwise, the use is similar with advantages.It seems that starting from v0.7 vpred works flawlessly now. It shouldn't suffer from ignorance of tags close to the 75tokens chunk borders like nai. It is more difficult to get burned images - even on cfg7 usually it just over-saturated but with smooth gradients, which can be useful for some styles. Yes it can make anything from (0,0,0) to (255,255,255). You will find brightness meta tags described above quite useful for easier/lazy prompting, natural text expressions also work. To get the most dark image - put high brightness into negative and/or use low brightness, low gamma tags. If you don't like very bright skin on dark background and want to reduce contrast (or on the contrary, enhance the effect) - use hdr/sdr in negative/positive.It was reported that in rare cases on some prompts there is a drop in contrast. Looks like other vpred models have same behaviour with such prompts, adding a \"separator\" closer to the border of the 75-token chunk fixes this. However, with 0.7 I haven't encountered this myself.To launch vpred version you will need dev build of A1111, Comfy (with special loader node), Forge or Reforge. Just use same parameters (Euler a, cfg 3..5, 20..28 steps) like epsilon. No need to use Cfg rescale, but you can try it, cfg++ works great.Base model:The model here has a small unet polishint after main training to improve small details, bump up resolution and others. Hovewer, you may be also interested into a RouWei-Base, which sometimes can perform better at complex prompts despite having minor mistakes in small details. It also comes in FP32, for example if you want to use fp32 text encoder nodes in Comfy, merge it or finetune.It can be found in Huggingface repo https://huggingface.co/Minthy/RouWei-0.8Known issues:Off course there are:Artists and style tags must be seperated into a different chunk from main prompt or come very lastThere may be some positional or combinational bias in rare cases, but it's not yet clear.There are some complaints about few of the general styles.Epsilon version relies too much on brightness meta tags, sometimes you will need to use them to get desired brightness shiftSome newly added styles/characters might be not as good and disctinct as they deserve toTo be discoveredRequests for artists/characters in future models are open. If you find artist/character/concept that perform weak, inaccurate or has strong watermark - please report, will add them explicitly. Follow for a new versions.JOIN THE DISCORD SERVER https://discord.gg/ZXHENAhqE9License:Same as illustrious. Fell free to use in your merges, finetunes, ets. but please leave a link or mention, it is mandatoryHow it's madeI'll consider to make a report or something like it later. For sure.In short, 98% of work is related to dataset preparations. Instead of blindly relying on loss-weighting based on tag frequency from nai paper, a custom guided loss-weighting implementation along with asynchronous collator for balancing have been used. Ztsnr (or close to it) with Epsilon prediction was achieved using noise scheduler augmentation.Spent compute - over 8k hours of H100 (apart from research and fail attempts)Thanks:First of all I'd like to acknowledge everyone who supports open source, develops in improves code. Thanks to the authors of illustrious for releasing model, thank to NoobAI team for being pioneers in open finetuning of such a scale, sharing experience, raising and solving issues that previously went unnoticed.Personal:Artists wish to remain anonymous for sharing private works; Few anonymous persons - donations, code, captions, etc., Soviet Cat - GPU sponsoring; Sv1. - llm access, captioning, code; K. - training code; Bakariso - datasets, testing, advices, insides; NeuroSenko - donations, testing, code; LOL2024 - a lot of unique datasets; T.,[] - datasets, testing, advises; rred, dga, Fi., ello - donations; TekeshiX - datasets. And other fellow brothers that helped. Love you so much \u2764\ufe0f.And off course everyone who made feedback and requests, it's really valuable.If I forgot to mention anyone, please notify.DonationsIf you want to support - share my models, leave feedback, make a cute picture with kemonomimi-girl. And of course, support original artists.AI is my hobby, I'm spending money on it and not begging for donations. However, it has turned into a large-scale and expensive undertaking. Consider to support to accelerate new training and researches.(Just keep in mind that I can waste it on alcohol or cosplay girls)BTC: bc1qwv83ggq8rvv07uk6dv4njs0j3yygj3aax4wg6cETH/USDT(e): 0x04C8a749F49aE8a56CB84cF0C99CD9E92eDB17dbXMR: 47F7JAyKP8tMBtzwxpoZsUVB8wzg2VrbtDKBice9FAS1FikbHEXXPof4PAb42CQ5ch8p8Hs4RvJuzPHDtaVSdQzD6ZbA5TZif you can offer gpu-time (a100+) - PM.",
     "sd version": "Other"
 }

Stable-diffusion/rouwei_v07Vpred.json

@@ -3,6 +3,6 @@
     "modelId": 950531,
     "modelVersionId": 1360303,
     "activation text": "",
-    "description": "Large scale finetune of Illustrious with state of the art performance.(tl/dr: Works exactly as it should without flaws you might encounter in other checkpoints.)V0.7 comes with the dataset of 7M images (~2M with natural text captions):New knowledge for characters, concepts, styles https://mega.nz/folder/lGQSVDyZ#uzuzfSlPRzgsWsxPQPbuhwMost of the things that was not good enough in 0.6 now on decent levelEven better prompting with good segmentation and natural phrases understanding (now it's even possible to easily make one character to cosplay other without typical issues 1, 2, 3)Better stability and broader coverage, less biasesExcelent performance for both very simple and complex promptsLots of general changes and fixesVpred version for v0.7 is released!Key advantages:Easy and convenient promptingGreat aesthetic, anatomy, stability along with versatilityVibrant colors and smooth gradients without trace of burningFull brightness range even with epsilon35k+ artist styles https://mega.nz/folder/lGQSVDyZ#uzuzfSlPRzgsWsxPQPbuhw, many general styles, almost any characterPure training without any merged weights, loras, tweakers, etc., you can add your own if neededAn addition to mentioned, comparing with vanilla Illustrious and NoobAI:No more annoying watermarksNo tags bleed and better segmentation (1, 2, 3 (nsfw))No characters bleed and related side effects (unwanted outfits, style, composition changes)No spawning of strange creatures, sfx on background or extra pair of breastsBetter coherence, prompt following, anatomy (significant boost over illustrious, slight over noob)Artist styles look exactly as they should (and lots of new added)Better prompt following without ignoring tags and need of (higher weights:1.4)Forget about long scizo-negativeStable style without random fluctuations on different seedsNew charactersDataset cut-off - 20th December 2024.Features and prompting:Basic:The checkpoint works both with short-simple and long-complex prompts. However, if there are contradictory or weird things - unlike with other checkpoints, they won't be ignored affecting the output. No guide-rails, no safeguards, no lobotomy.Just prompt what you want to see and don't prompt what shouldn't be on the picture. If you want to have a view from above - don't put ceiling into positive, if you want to have crop view with head out of frame - don't make detailed description of character facial features, and so on. Pretty simple but sometimes people are missing it.Version 0.7 comes with several improvements in prompt understanding and segmentation.However, because SDXL is based on CLIP text encoders which have limit of 75 (77) tokens for input, it is quite limited due to prompt chunking. If you want to specify some features for character/object and separate them from other prompt parts - make sure they are in the same chunk and optionally separate it with BREAK.It will not solve problem of traits mixing completely, but can reduce it improving overall understanding, since text encoders on RouWei are able to process the whole sequence, not individual concepts better then others.Dataset contains only booru-style tags and (simplified) natural text expressions. Despite having a share of furries, all captions have been converted to classic booru style to avoid a number of problems that may arise when mixing different systems. So e621 tags won't be understanded properly.Sampling parameters:~1 megapixel for txt2img, any AR with resolution multiple of 32 (1024x1024, 1056x, 1152x, 1216x832,...). Euler_a, 20..28steps.CFG: for epsilon version 4..9 (7 is best), for vpred version, 3..5Sigmas multiply may improve results a bit, CFG++ samplers work fine. LCM/PCM/DMD/... and exotic samplers untested.Karras scheduler doesn't work well, same for some samplers.Highresfix - x1.5 latent + denoise 0.6 or any gan + denoise 0.3..0.55.For vpred version lower CFG 3..5 is needed!For vpred version lower CFG 3..5 is needed!Quality classification:Only 4 quality tags:masterpiece, best qualitylow quality, worst qualityNothing else. Meta tags like lowres have been removed, better not to use them. Low resolution images have been either removed or upscaled and cleaned with DAT depending on their importance.Negative prompt:worst quality, low quality, watermarkThat's all, no need of \"rusty trombone\", \"farting on prey\" and others. Do not put tags like greyscale, monochrome in negative unless you understand what are you doing. It will lead to burning and over-saturation, colors are fine out of box.Hovewer if you want to boost or adjust it - please use extra tags from brightness/colors/contrast section below.Artist styles:Grids with examples https://mega.nz/folder/lGQSVDyZ#uzuzfSlPRzgsWsxPQPbuhw, list https://files.catbox.moe/q6q80c.txt (also can be found in \"training data\").Used with \"by \" it's mandatory. It will not work properly without it.\"by \" is used as meta-token for styles to avoid mixing/misinterpret with tags/characters of similar or close name. This allows to have a better results for styles and at the same time avoid random style fluctuation that you may observe in some other checkpoints.Multiple give very interesting results, can be controlled with prompt weights.General styles:2.5d, anime screencap, bold line, sketch, cgi, digital painting, flat colors, smooth shading, minimalistic, ink style, oil style, pastel styleBooru tags styles:1950s (style), 1960s (style), 1970s (style), 1980s (style), 1990s (style), 1990s (style), animification, art nouveau, pinup (style), toon (style), western comics (style), nihonga, shikishi, minimalism, fine art parodyand everything from this group https://danbooru.donmai.us/wiki_pages/traditional_media.Can be used in combinations (with artists too), with weights, both in positive and negative prompts.Characters:Use full name booru tag and proper formatting, like karin_(blue_archive) -> karin \\(blue_archive\\), use skin tags for better reproducing, like karin \\(bunny \\(blue_archive\\). Autocomplete extension might be very useful.Most characters are recognized just by their booru tag, but it will be more accurate if you describe their basic traits. Here you can easily redress your waifu/husbendo just by the prompt without suffering from the typical leaks of basic features.Natural text:Use it in combination with booru tags, works great. Use only natural text after typing styles and quality tags. Use just booru tags and forget about it, it's all up to you. To get best performance keep track if CLIP 75 tokens chunks.About 2M images in dataset have hybrid natural-text captions made by Opus-Vision, GPT-4o and ToriiGate https://huggingface.co/Minthy/ToriiGate-v0.4-7B, which improves segmentation, allows to get some understanding and provides a response to chain of phrases, not single words. But overall performance is not even close to flux or sd3.5.Lots of Tail/Ears-related concepts:tail censor, holding own tail, hugging own tail, holding another's tail, tail grab, tail raised, tail down, ears down, hand on own ear, tail around own leg, tail around penis, tail through clothes, tail under clothes, lifted by tail, tail biting, tail insertion, tail masturbation, holding with tail, ...(booru meaning, not e621) and many others with natural text. The majority works perfectly, some requires rolling.Brightness/colors/contrast:You can use extra meta tags to control it:low brightness, high brightness, low saturation, high saturation, low gamma, high gamma, sharp colors, soft colors, hdr, sdrExampleThey work both in epsilon and vpred version and works really good.Unfortunately here is an issue - the model (epsilon) relies on them too much. Without low brightness or low gamma or limited range (in negative) it might be difficult to achieve true 0,0,0 black, the same often true for white.Both epsilon and vpred versions have like true zsnr, full range of colors and brightness without common flaws observed. But they behaves differently, just try it.Vpred versionMain thing you need to know - lower your CFG from 7 down to 5 (or less). Otherwise, the use is similar with advantages.It seems that from v0.7 vpred works flawlessly now. It shouldn't suffer from ignorance of tags close to the 75tokens chunk borders like nai. It is more difficult to get burned images - even on cfg7 usually it just over-saturated but with smooth gradients, which can be usefull for some styles. Yes it can make anything from (0,0,0) to (255,255,255). You will find brightness meta tags described above quite useful for easier/lazy prompting. To get the most dark image - put high brightness into negative and/or use low brightness, low gamma tags. If you don't like very bright skin on dark background and want to reduce contrast (or on the contrary, enhance the effect) - use hdr/sdr in negative/posive.It was reported that in rare cases on some promts there is a drop in contrast. Looks like other vpred models have same behavior with such prompts, adding a \"separator\" closer to the border of the 75-token chunk fixes this. However, with 0.7 I haven't encountered this myself.To launch vpred version you will need dev build of A1111, comfy (with special loader node) or Reforge. Just use same parameters (Euler a, cfg 3..5, 20..28 steps) like epsilon. No need to use Cfg rescale, but you can try it.Base model:Epsilon and vpred versions here have a brief aesthetic polishing after main training to improve small details and coherence. If you want to use RouWei in merges, extract something without bringing that last things, or finetune it - you can find base version of RouWei on huggingface https://huggingface.co/Minthy/RouWei-0.7.Known issues:Off course there are:Epsilon version relies too much on brightness meta tags, sometimes you will need to use them to get desired brightness shiftSome newly added styles/characters might be not as good and disctinct as they deserve toInferior in furry-related knowledge compared to NoobAiTo be discoveredRequests for artists/characters in future models are open. If you find artist/character/concept that perform weak, inaccurate or has strong watermark - please report, will add them explicitly. Follow for a new versions.JOIN THE DISCORD SERVER https://discord.gg/ZXHENAhqE9License:Same as illustrious. Fell free to use in your merges, finetunes, ets. just please leave a link.How it's madeI'll consider to make a report or something like it later. For sure.In short, 98% of work is related to dataset preparations. Instead of blindly relying on loss-weighting based on tag frequency from nai paper, a custom guided loss-weighting implementation along with asynchronous collator for balancing have been used. Ztsnr (or close to it) with Epsilon prediction was achieved using noise scheduler augmentation.Spent compute - about 25days of 4xH100 (apart from research and fail attempts)Thanks:First of all I'd like to acknowledge everyone who supports open source, develops in improves code. Thanks to the authors of illustrious for releasing model, thank to NoobAI team for being pioneers in open finetuning of such a scale, sharing experience, raising and solving issues that previously went unnoticed.Personal:Artists wish to remain anonymous for sharing private works; Few anonymous persons - donations, code, captions, etc., Soviet Cat - GPU sponsoring; Sv1. - llm access, captioning, code; K. - training code; Bakariso - datasets, testing, advices, insides; NeuroSenko - donations, testing, code; LOL2024 - a lot of unique datasets; T.,[] - datasets, testing, advises; rred, dga, Fi., ello - donations; other fellow brothers that helped. Love you so much \u2764\ufe0f.And off course everyone who made feedback and requests, it's really valuable.If I forgot to mention anyone, please notify.DonationsIf you want to support - share my models, leave feedback, make a cute picture with kemonomimi-girl. And of course, support original artists.AI is my hobby, I'm spending money on it and not begging for donations. However, it has turned into a large-scale and expensive undertaking. Consider to support to accelerate new training and researches.(Just keep in mind that I can waste it on alcohol or cosplay girls)BTC: bc1qwv83ggq8rvv07uk6dv4njs0j3yygj3aax4wg6cETH/USDT(e): 0x04C8a749F49aE8a56CB84cF0C99CD9E92eDB17dbXMR: 47F7JAyKP8tMBtzwxpoZsUVB8wzg2VrbtDKBice9FAS1FikbHEXXPof4PAb42CQ5ch8p8Hs4RvJuzPHDtaVSdQzD6ZbA5TZif you can offer gpu-time (a100+) - PM.",
+    "description": "In depth retraining of Illustrious to achieve best prompt adherence, knowledge and state of the art performance.Big dreams come trueThe version number is just an index of current final release, not a fraction of the planned training.HF repo https://huggingface.co/Minthy/RouWei-0.8Large scale finetune using gpu cluster with a dataset of ~13M pictures (~4M with natural text captions)Fresh and wast knowledge about characters, concepts, styles, cultural and related thingsThe best prompt adherence among SDXL anime models at the moment of releaseSolved main problems with tags bleeding and biases, common for Illustrious, NoobAi and other checkpointsExcellent aesthetics and knowledge across a wide range of styles (over 50,000 artists https://files.catbox.moe/anleso.txt (examples https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg), including hundreds of unique cherry-picked datasets from private galleries, including those received from the artists themselves)High flexibility and variety without stability tradeoffNo more annoying watermarks for popular styles thanks to clean datasetVibrant colors and smooth gradients without trace of burning, full range even with epsilonPure training from Illustrious v0.1 without involving third-party checkpoints, Loras, tweakers, etc.There are also some issues and changes compared to the previous version, please RTFM.Dataset cut-off - end of April 2025.Features and prompting:Important change:When you are prompting artist styles, especially mixing several, their tags MUST BE in a separate CLIP chunk. Just add BREAK after it (for A1111 and derivatives), use conditioning concat node (for Comfy) or at least put them in the very end. Otherwise, significant degradation of results is likely.Basic:The checkpoint works both with short-simple and long-complex prompts. However, if there are contradictory or weird things - unlike with others they won't be ignored affecting the output. No guide-rails, no safeguards, no lobotomy.Just prompt what you want to see and don't prompt what shouldn't be on the picture. If you want to have a view from above - don't put ceiling into positive, if you want to have crop view with head out of frame - don't make detailed description of character facial features, and so on. Pretty simple but sometimes people are missing it.Version 0.8 comes with advanced understanding of natural text prompts. It doesn't mean that you are obligated to use it, tags only - completely fine, especially because understanding of tags combinations is also improved.Do not expect it to perform like Flux or other models based on T5 or LLM text encoders. The whole size ot SDXL checkpoint is less then only that text encoder, in addition illustrious-v0.1 which is used as the base completely forgot a lot of general things from vanilla sdxl-base.However, even in current state it works much better, allows to do new things usually impossible without external guidance, as well making manual editing, inpainting, etc more convenient.To achieve best performance you should keep track of CLIP chunks. In SDXL the prompt is separated into a chunks of 75 (77 including BOS and EOS) tokens, that are processing by CLIP separately, and only then are concatinating and comes as conditions to unet.If you want to specify some features for character/object and separate them from other prompt parts - make sure they are in the same chunk and optionally separate it with BREAK. It will not solve problem of traits mixing completely, but can reduce it improving overall understanding, since text encoders on RouWei are able to process the whole sequence, not individual concepts better then others.Dataset contains only booru-style tags and natural text expressions. Despite having a share of furries, real life photos, western media, etc. all captions have been converted to classic booru style to avoid a number of problems from mixing of different systems. So e621 tags won't be understanded properly.Sampling parameters:~1 megapixel for txt2img, any AR with resolution multiple of 32 (1024x1024, 1056x, 1152x, 1216x832,...). Euler_a, 20..28steps.CFG: for epsilon version 4..9 (7 is best), for vpred version, 3..5Sigmas multiply may improve results a bit, CFG++ samplers work fine. LCM/PCM/DMD/... and exotic samplers untested.Some schedulers doesn't work well.Highresfix - x1.5 latent + denoise 0.6 or any gan + denoise 0.3..0.55.For vpred version lower CFG 3..5 is needed!For vpred version lower CFG 3..5 is needed!Quality classification:Only 4 quality tags:masterpiece, best qualityfor positive andlow quality, worst qualityfor negative.Nothing else. Actually you can even omit positive and reduce negative to low quality only, since they can affect basic style and composition.Meta tags like lowres have been removed and don't work, better not to use them. Low resolution images have been either removed or upscaled and cleaned with DAT depending on their importance.Negative prompt:worst quality, low quality, watermarkThat's all, no need of \"rusty trombone\", \"farting on prey\" and others. Do not put tags like greyscale, monochrome in negative unless you understand what are you doing. Extra tags for brightness/colors/contrast section below can be usedArtist styles:Grids with examples https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg, list/wildcard https://files.catbox.moe/anleso.txt (also can be found in \"training data\").Used with \"by \" it's mandatory. It will not work properly without it.\"by \" is a meta-token for styles to avoid mixing/misinterpret with tags/characters of similar or close name. This allows to have a better results for styles and at the same time avoid random style fluctuation that you may observe in other checkpoints.Multiple give very interesting results, can be controlled with prompt weights and spells.YOU MUST ADD BREAK after artists/style tags (for A1111) or concat conditioning (for Comfy) or put them in the very end of your prompt.For example:by kantoku, by wlop, best quality, masterpiece BREAK 1girl, ...General styles:2.5d, anime screencap, bold line, sketch, cgi, digital painting, flat colors, smooth shading, minimalistic, ink style, oil style, pastel styleBooru tags styles:1950s (style), 1960s (style), 1970s (style), 1980s (style), 1990s (style), 2000s (style), animification, art nouveau, pinup (style), toon (style), western comics (style), nihonga, shikishi, minimalism, fine art parodyand everything from this group https://danbooru.donmai.us/wiki_pages/traditional_media.Can be used in combinations (with artists too), with weights, both in positive and negative prompts.Characters:Use full name booru tag and proper formatting, like karin_(blue_archive) -> karin \\(blue archive\\), use skin tags for better reproducing, like karin \\(bunny\\) \\(blue archive\\). Autocomplete extension might be very useful.Most characters are recognized just by their booru tag, but it will be more accurate if you describe their basic traits. Here you can easily redress your waifu/husbendo just by the prompt without suffering from the typical leaks of basic features.Natural text:Use it in combination with booru tags, works great. Use only natural text after typing styles and quality tags. Use just booru tags and forget about it, it's all up to you. To get best performance keep track if CLIP 75 tokens chunks.About 4M of images in dataset had hybrid natural-text captions, made by Claude, GPT, Gemini, ToriiGate https://huggingface.co/Minthy/ToriiGate-v0.4-7B, then refactored, cleaned and combined with tags in different variations for augmentation.Unlike typical captions, these contains character names which is very useful. Better to keep it clean, short and convenient description works best. Better not use long and sloppy BS likeA mysteriously enchanting feminine entity of indeterminate yet youthful essence, whose celestial visage radiates with the ethereal luminescence of a thousand dying stars, blessed with locks cascading like the golden rivers of ancient mythology, perhaps styled in a manner reminiscent of contemporary fashion trends though not necessarily adhering to any specific aesthetic paradigm. Her eyes, pools of unfathomable depth and hue, sparkle with the wisdom of millennia yet maintain an innocent quality that defies temporal constraints...For captioning you can use ToriiGate https://huggingface.co/Minthy/ToriiGate-v0.4-7B in short mode.And don't expect it to be as good as flux and others, it tries very hard and after several rolls usually you can get what you want, but it is not that stable and detailed.Lots of Tail/Ears-related concepts:Oh yeahtail censor, holding own tail, hugging own tail, holding another's tail, tail grab, tail raised, tail down, ears down, hand on own ear, tail around own leg, tail around penis, tailjob, tail through clothes, tail under clothes, lifted by tail, tail biting, tail penetration (including a specific indication of vaginal/anal), tail masturbation, holding with tail, panties on tail, bra on tail, tail focus, presenting own tail...(booru meaning, not e621) and many others with natural text. The majority works perfectly, some requires a lot of rolling.Brightness/colors/contrast:You can use extra meta tags to control it:low brightness, high brightness, low saturation, high saturation, low gamma, high gamma, sharp colors, soft colors, hdr, sdrExampleThey work both in epsilon and vpred version and works really good.Epsilon version relies on them too much. Without low brightness or low gamma or limited range (in negative) it might be difficult to achieve true 0,0,0 black, the same often true for white.Both epsilon and vpred versions have like true zsnr, full range of colors and brightness without common flaws observed. But they behaves differently, just try it.Vpred versionMain thing you need to know - lower your CFG from 7 down to 5 (or less). Otherwise, the use is similar with advantages.It seems that starting from v0.7 vpred works flawlessly now. It shouldn't suffer from ignorance of tags close to the 75tokens chunk borders like nai. It is more difficult to get burned images - even on cfg7 usually it just over-saturated but with smooth gradients, which can be useful for some styles. Yes it can make anything from (0,0,0) to (255,255,255). You will find brightness meta tags described above quite useful for easier/lazy prompting, natural text expressions also work. To get the most dark image - put high brightness into negative and/or use low brightness, low gamma tags. If you don't like very bright skin on dark background and want to reduce contrast (or on the contrary, enhance the effect) - use hdr/sdr in negative/positive.It was reported that in rare cases on some prompts there is a drop in contrast. Looks like other vpred models have same behaviour with such prompts, adding a \"separator\" closer to the border of the 75-token chunk fixes this. However, with 0.7 I haven't encountered this myself.To launch vpred version you will need dev build of A1111, Comfy (with special loader node), Forge or Reforge. Just use same parameters (Euler a, cfg 3..5, 20..28 steps) like epsilon. No need to use Cfg rescale, but you can try it, cfg++ works great.Base model:The model here has a small unet polishint after main training to improve small details, bump up resolution and others. Hovewer, you may be also interested into a RouWei-Base, which sometimes can perform better at complex prompts despite having minor mistakes in small details. It also comes in FP32, for example if you want to use fp32 text encoder nodes in Comfy, merge it or finetune.It can be found in Huggingface repo https://huggingface.co/Minthy/RouWei-0.8Known issues:Off course there are:Artists and style tags must be seperated into a different chunk from main prompt or come very lastThere may be some positional or combinational bias in rare cases, but it's not yet clear.There are some complaints about few of the general styles.Epsilon version relies too much on brightness meta tags, sometimes you will need to use them to get desired brightness shiftSome newly added styles/characters might be not as good and disctinct as they deserve toTo be discoveredRequests for artists/characters in future models are open. If you find artist/character/concept that perform weak, inaccurate or has strong watermark - please report, will add them explicitly. Follow for a new versions.JOIN THE DISCORD SERVER https://discord.gg/ZXHENAhqE9License:Same as illustrious. Fell free to use in your merges, finetunes, ets. but please leave a link or mention, it is mandatoryHow it's madeI'll consider to make a report or something like it later. For sure.In short, 98% of work is related to dataset preparations. Instead of blindly relying on loss-weighting based on tag frequency from nai paper, a custom guided loss-weighting implementation along with asynchronous collator for balancing have been used. Ztsnr (or close to it) with Epsilon prediction was achieved using noise scheduler augmentation.Spent compute - over 8k hours of H100 (apart from research and fail attempts)Thanks:First of all I'd like to acknowledge everyone who supports open source, develops in improves code. Thanks to the authors of illustrious for releasing model, thank to NoobAI team for being pioneers in open finetuning of such a scale, sharing experience, raising and solving issues that previously went unnoticed.Personal:Artists wish to remain anonymous for sharing private works; Few anonymous persons - donations, code, captions, etc., Soviet Cat - GPU sponsoring; Sv1. - llm access, captioning, code; K. - training code; Bakariso - datasets, testing, advices, insides; NeuroSenko - donations, testing, code; LOL2024 - a lot of unique datasets; T.,[] - datasets, testing, advises; rred, dga, Fi., ello - donations; TekeshiX - datasets. And other fellow brothers that helped. Love you so much \u2764\ufe0f.And off course everyone who made feedback and requests, it's really valuable.If I forgot to mention anyone, please notify.DonationsIf you want to support - share my models, leave feedback, make a cute picture with kemonomimi-girl. And of course, support original artists.AI is my hobby, I'm spending money on it and not begging for donations. However, it has turned into a large-scale and expensive undertaking. Consider to support to accelerate new training and researches.(Just keep in mind that I can waste it on alcohol or cosplay girls)BTC: bc1qwv83ggq8rvv07uk6dv4njs0j3yygj3aax4wg6cETH/USDT(e): 0x04C8a749F49aE8a56CB84cF0C99CD9E92eDB17dbXMR: 47F7JAyKP8tMBtzwxpoZsUVB8wzg2VrbtDKBice9FAS1FikbHEXXPof4PAb42CQ5ch8p8Hs4RvJuzPHDtaVSdQzD6ZbA5TZif you can offer gpu-time (a100+) - PM.",
     "sd version": "Other"
 }

Stable-diffusion/rouwei_v080Epsilon.json

@@ -3,6 +3,6 @@
     "modelId": 950531,
     "modelVersionId": 1832460,
     "activation text": "",
-    "description": "In depth retraining of Illustrious to achieve best prompt adherence, knowledge and state of the art performance.Big dreams come trueThe version number is just an index of current final release, not a fraction of the planned training.HF repo https://huggingface.co/Minthy/RouWei-0.8Large scale finetune using gpu cluster with a dataset of ~13M pictures (~4M with natural text captions)Fresh and wast knowledge about characters, concepts, styles, cultural and related thingsThe best prompt adherence among SDXL anime models at the moment of releaseSolved main problems with tags bleeding and biases, common for Illustrious, NoobAi and other checkpointsExcellent aesthetics and knowledge across a wide range of styles (over 50,000 artists https://files.catbox.moe/anleso.txt, including hundreds of unique cherry-picked datasets from private galleries, including those received from the artists themselves)High flexibility and variety without stability tradeoffNo more annoying watermarks for popular styles thanks to clean datasetVibrant colors and smooth gradients without trace of burning, full range even with epsilonPure training from Illustrious v0.1 without involving third-party checkpoints, Loras, tweakers, etc.There are also some issues and changes compared to the previous version, please RTFM.Vpred version for v0.8 is baking, will be soon.Dataset cut-off - end of April 2025.Features and prompting:Important change:When you are prompting artist styles, especially mixing several, their tags MUST BE in a separate CLIP chunk. Just add BREAK after it (for A1111 and derivatives), use conditioning concat node (for Comfy) or at least put them in the very end. Otherwise, significant degradation of results is likely.Basic:The checkpoint works both with short-simple and long-complex prompts. However, if there are contradictory or weird things - unlike with others they won't be ignored affecting the output. No guide-rails, no safeguards, no lobotomy.Just prompt what you want to see and don't prompt what shouldn't be on the picture. If you want to have a view from above - don't put ceiling into positive, if you want to have crop view with head out of frame - don't make detailed description of character facial features, and so on. Pretty simple but sometimes people are missing it.Version 0.8 comes with advanced understanding of natural text prompts. It doesn't mean that you are obligated to use it, tags only - completely fine, especially because understanding of tags combinations is also improved.Do not expect it to perform like Flux or other models based on T5 or LLM text encoders. The whole size ot SDXL checkpoint is less then only that text encoder, in addition illustrious-v0.1 which is used as the base completely forgot a lot of general things from vanilla sdxl-base.However, even in current state it works much better, allows to do new things usually impossible without external guidance, as well making manual editing, inpainting, etc more convenient.To achieve best performance you should keep track of CLIP chunks. In SDXL the prompt is separated into a chunks of 75 (77 including BOS and EOS) tokens, that are processing by CLIP separately, and only then are concatinating and comes as conditions to unet.If you want to specify some features for character/object and separate them from other prompt parts - make sure they are in the same chunk and optionally separate it with BREAK. It will not solve problem of traits mixing completely, but can reduce it improving overall understanding, since text encoders on RouWei are able to process the whole sequence, not individual concepts better then others.Dataset contains only booru-style tags and natural text expressions. Despite having a share of furries, real life photos, western media, etc. all captions have been converted to classic booru style to avoid a number of problems from mixing of different systems. So e621 tags won't be understanded properly.Sampling parameters:~1 megapixel for txt2img, any AR with resolution multiple of 32 (1024x1024, 1056x, 1152x, 1216x832,...). Euler_a, 20..28steps.CFG: for epsilon version 4..9 (7 is best), for vpred version, 3..5Sigmas multiply may improve results a bit, CFG++ samplers work fine. LCM/PCM/DMD/... and exotic samplers untested.Some schedulers doesn't work well.Highresfix - x1.5 latent + denoise 0.6 or any gan + denoise 0.3..0.55.For vpred version lower CFG 3..5 is needed!For vpred version lower CFG 3..5 is needed!Quality classification:Only 4 quality tags:masterpiece, best qualityfor positive andlow quality, worst qualityfor negative.Nothing else. Actually you can even omit positive and reduce negative to low quality only, since they can affect basic style and composition.Meta tags like lowres have been removed and don't work, better not to use them. Low resolution images have been either removed or upscaled and cleaned with DAT depending on their importance.Negative prompt:worst quality, low quality, watermarkThat's all, no need of \"rusty trombone\", \"farting on prey\" and others. Do not put tags like greyscale, monochrome in negative unless you understand what are you doing. Extra tags for brightness/colors/contrast section below can be usedArtist styles:Grids with examples (comming soon), list/wildcard https://files.catbox.moe/anleso.txt (also can be found in \"training data\").Used with \"by \" it's mandatory. It will not work properly without it.\"by \" is a meta-token for styles to avoid mixing/misinterpret with tags/characters of similar or close name. This allows to have a better results for styles and at the same time avoid random style fluctuation that you may observe in other checkpoints.Multiple give very interesting results, can be controlled with prompt weights and spells.YOU MUST ADD BREAK after artists/style tags (for A1111) or concat conditioning (for Comfy) or put them in the very end of your prompt.For example:by kantoku, by wlop, best quality, masterpiece BREAK 1girl, ...General styles:2.5d, anime screencap, bold line, sketch, cgi, digital painting, flat colors, smooth shading, minimalistic, ink style, oil style, pastel styleBooru tags styles:1950s (style), 1960s (style), 1970s (style), 1980s (style), 1990s (style), 2000s (style), animification, art nouveau, pinup (style), toon (style), western comics (style), nihonga, shikishi, minimalism, fine art parodyand everything from this group https://danbooru.donmai.us/wiki_pages/traditional_media.Can be used in combinations (with artists too), with weights, both in positive and negative prompts.Characters:Use full name booru tag and proper formatting, like karin_(blue_archive) -> karin \\(blue archive\\), use skin tags for better reproducing, like karin \\(bunny\\) \\(blue archive\\). Autocomplete extension might be very useful.Most characters are recognized just by their booru tag, but it will be more accurate if you describe their basic traits. Here you can easily redress your waifu/husbendo just by the prompt without suffering from the typical leaks of basic features.Natural text:Use it in combination with booru tags, works great. Use only natural text after typing styles and quality tags. Use just booru tags and forget about it, it's all up to you. To get best performance keep track if CLIP 75 tokens chunks.About 4M of images in dataset had hybrid natural-text captions, made by Claude, GPT, Gemini, ToriiGate https://huggingface.co/Minthy/ToriiGate-v0.4-7B, then refactored, cleaned and combined with tags in different variations for augmentation.Unlike typical captions, these contains character names which is very useful. Better to keep it clean, short and convenient description works best. Better not use long and sloppy BS likeA mysteriously enchanting feminine entity of indeterminate yet youthful essence, whose celestial visage radiates with the ethereal luminescence of a thousand dying stars, blessed with locks cascading like the golden rivers of ancient mythology, perhaps styled in a manner reminiscent of contemporary fashion trends though not necessarily adhering to any specific aesthetic paradigm. Her eyes, pools of unfathomable depth and hue, sparkle with the wisdom of millennia yet maintain an innocent quality that defies temporal constraints...For captioning you can use ToriiGate https://huggingface.co/Minthy/ToriiGate-v0.4-7B in short mode.And don't expect it to be as good as flux and others, it tries very hard and after several rolls usually you can get what you want, but it is not that stable and detailed.Lots of Tail/Ears-related concepts:Oh yeahtail censor, holding own tail, hugging own tail, holding another's tail, tail grab, tail raised, tail down, ears down, hand on own ear, tail around own leg, tail around penis, tailjob, tail through clothes, tail under clothes, lifted by tail, tail biting, tail penetration (including a specific indication of vaginal/anal), tail masturbation, holding with tail, panties on tail, bra on tail, tail focus, presenting own tail...(booru meaning, not e621) and many others with natural text. The majority works perfectly, some requires a lot of rolling.Brightness/colors/contrast:You can use extra meta tags to control it:low brightness, high brightness, low saturation, high saturation, low gamma, high gamma, sharp colors, soft colors, hdr, sdrExampleThey work both in epsilon and vpred version and works really good.Epsilon version relies on them too much. Without low brightness or low gamma or limited range (in negative) it might be difficult to achieve true 0,0,0 black, the same often true for white.Both epsilon and vpred versions have like true zsnr, full range of colors and brightness without common flaws observed. But they behaves differently, just try it.Vpred versionVpred for v0.8 is coming soon. Info below is related to v0.7Main thing you need to know - lower your CFG from 7 down to 5 (or less). Otherwise, the use is similar with advantages.It seems that from v0.7 vpred works flawlessly now. It shouldn't suffer from ignorance of tags close to the 75tokens chunk borders like nai. It is more difficult to get burned images - even on cfg7 usually it just over-saturated but with smooth gradients, which can be usefull for some styles. Yes it can make anything from (0,0,0) to (255,255,255). You will find brightness meta tags described above quite useful for easier/lazy prompting. To get the most dark image - put high brightness into negative and/or use low brightness, low gamma tags. If you don't like very bright skin on dark background and want to reduce contrast (or on the contrary, enhance the effect) - use hdr/sdr in negative/posive.It was reported that in rare cases on some promts there is a drop in contrast. Looks like other vpred models have same behavior with such prompts, adding a \"separator\" closer to the border of the 75-token chunk fixes this. However, with 0.7 I haven't encountered this myself.To launch vpred version you will need dev build of A1111, Comfy (with special loader node), Forge or Reforge. Just use same parameters (Euler a, cfg 3..5, 20..28 steps) like epsilon. No need to use Cfg rescale, but you can try it, cfg++ works great.Base model:The model here has a small unet polishint after main training to improve small details, bump up resolution and others. Hovewer, you may be also interested into a RouWei-Base, which sometimes can perform better at complex prompts despite having minor mistakes in small details. It also comes in FP32, for example if you want to use fp32 text encoder nodes in Comfy, merge it or finetune.It can be found in Huggingface repo https://huggingface.co/Minthy/RouWei-0.8Known issues:Off course there are:Artists and style tags must be seperated into a different chunk from main prompt or come very lastThere may be some positional or combinational bias in rare cases, but it's not yet clear.There are some complaints about few of the general styles.Epsilon version relies too much on brightness meta tags, sometimes you will need to use them to get desired brightness shiftSome newly added styles/characters might be not as good and disctinct as they deserve toTo be discoveredRequests for artists/characters in future models are open. If you find artist/character/concept that perform weak, inaccurate or has strong watermark - please report, will add them explicitly. Follow for a new versions.JOIN THE DISCORD SERVER https://discord.gg/ZXHENAhqE9License:Same as illustrious. Fell free to use in your merges, finetunes, ets. but please leave a link or mention, it is mandatoryHow it's madeI'll consider to make a report or something like it later. For sure.In short, 98% of work is related to dataset preparations. Instead of blindly relying on loss-weighting based on tag frequency from nai paper, a custom guided loss-weighting implementation along with asynchronous collator for balancing have been used. Ztsnr (or close to it) with Epsilon prediction was achieved using noise scheduler augmentation.Spent compute - over 8k hours of H100 (apart from research and fail attempts)Thanks:First of all I'd like to acknowledge everyone who supports open source, develops in improves code. Thanks to the authors of illustrious for releasing model, thank to NoobAI team for being pioneers in open finetuning of such a scale, sharing experience, raising and solving issues that previously went unnoticed.Personal:Artists wish to remain anonymous for sharing private works; Few anonymous persons - donations, code, captions, etc., Soviet Cat - GPU sponsoring; Sv1. - llm access, captioning, code; K. - training code; Bakariso - datasets, testing, advices, insides; NeuroSenko - donations, testing, code; LOL2024 - a lot of unique datasets; T.,[] - datasets, testing, advises; rred, dga, Fi., ello - donations; TekeshiX - datasets. And other fellow brothers that helped. Love you so much \u2764\ufe0f.And off course everyone who made feedback and requests, it's really valuable.If I forgot to mention anyone, please notify.DonationsIf you want to support - share my models, leave feedback, make a cute picture with kemonomimi-girl. And of course, support original artists.AI is my hobby, I'm spending money on it and not begging for donations. However, it has turned into a large-scale and expensive undertaking. Consider to support to accelerate new training and researches.(Just keep in mind that I can waste it on alcohol or cosplay girls)BTC: bc1qwv83ggq8rvv07uk6dv4njs0j3yygj3aax4wg6cETH/USDT(e): 0x04C8a749F49aE8a56CB84cF0C99CD9E92eDB17dbXMR: 47F7JAyKP8tMBtzwxpoZsUVB8wzg2VrbtDKBice9FAS1FikbHEXXPof4PAb42CQ5ch8p8Hs4RvJuzPHDtaVSdQzD6ZbA5TZif you can offer gpu-time (a100+) - PM.",
+    "description": "In depth retraining of Illustrious to achieve best prompt adherence, knowledge and state of the art performance.Big dreams come trueThe version number is just an index of current final release, not a fraction of the planned training.HF repo https://huggingface.co/Minthy/RouWei-0.8Large scale finetune using gpu cluster with a dataset of ~13M pictures (~4M with natural text captions)Fresh and wast knowledge about characters, concepts, styles, cultural and related thingsThe best prompt adherence among SDXL anime models at the moment of releaseSolved main problems with tags bleeding and biases, common for Illustrious, NoobAi and other checkpointsExcellent aesthetics and knowledge across a wide range of styles (over 50,000 artists https://files.catbox.moe/anleso.txt (examples https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg), including hundreds of unique cherry-picked datasets from private galleries, including those received from the artists themselves)High flexibility and variety without stability tradeoffNo more annoying watermarks for popular styles thanks to clean datasetVibrant colors and smooth gradients without trace of burning, full range even with epsilonPure training from Illustrious v0.1 without involving third-party checkpoints, Loras, tweakers, etc.There are also some issues and changes compared to the previous version, please RTFM.Dataset cut-off - end of April 2025.Features and prompting:Important change:When you are prompting artist styles, especially mixing several, their tags MUST BE in a separate CLIP chunk. Just add BREAK after it (for A1111 and derivatives), use conditioning concat node (for Comfy) or at least put them in the very end. Otherwise, significant degradation of results is likely.Basic:The checkpoint works both with short-simple and long-complex prompts. However, if there are contradictory or weird things - unlike with others they won't be ignored affecting the output. No guide-rails, no safeguards, no lobotomy.Just prompt what you want to see and don't prompt what shouldn't be on the picture. If you want to have a view from above - don't put ceiling into positive, if you want to have crop view with head out of frame - don't make detailed description of character facial features, and so on. Pretty simple but sometimes people are missing it.Version 0.8 comes with advanced understanding of natural text prompts. It doesn't mean that you are obligated to use it, tags only - completely fine, especially because understanding of tags combinations is also improved.Do not expect it to perform like Flux or other models based on T5 or LLM text encoders. The whole size ot SDXL checkpoint is less then only that text encoder, in addition illustrious-v0.1 which is used as the base completely forgot a lot of general things from vanilla sdxl-base.However, even in current state it works much better, allows to do new things usually impossible without external guidance, as well making manual editing, inpainting, etc more convenient.To achieve best performance you should keep track of CLIP chunks. In SDXL the prompt is separated into a chunks of 75 (77 including BOS and EOS) tokens, that are processing by CLIP separately, and only then are concatinating and comes as conditions to unet.If you want to specify some features for character/object and separate them from other prompt parts - make sure they are in the same chunk and optionally separate it with BREAK. It will not solve problem of traits mixing completely, but can reduce it improving overall understanding, since text encoders on RouWei are able to process the whole sequence, not individual concepts better then others.Dataset contains only booru-style tags and natural text expressions. Despite having a share of furries, real life photos, western media, etc. all captions have been converted to classic booru style to avoid a number of problems from mixing of different systems. So e621 tags won't be understanded properly.Sampling parameters:~1 megapixel for txt2img, any AR with resolution multiple of 32 (1024x1024, 1056x, 1152x, 1216x832,...). Euler_a, 20..28steps.CFG: for epsilon version 4..9 (7 is best), for vpred version, 3..5Sigmas multiply may improve results a bit, CFG++ samplers work fine. LCM/PCM/DMD/... and exotic samplers untested.Some schedulers doesn't work well.Highresfix - x1.5 latent + denoise 0.6 or any gan + denoise 0.3..0.55.For vpred version lower CFG 3..5 is needed!For vpred version lower CFG 3..5 is needed!Quality classification:Only 4 quality tags:masterpiece, best qualityfor positive andlow quality, worst qualityfor negative.Nothing else. Actually you can even omit positive and reduce negative to low quality only, since they can affect basic style and composition.Meta tags like lowres have been removed and don't work, better not to use them. Low resolution images have been either removed or upscaled and cleaned with DAT depending on their importance.Negative prompt:worst quality, low quality, watermarkThat's all, no need of \"rusty trombone\", \"farting on prey\" and others. Do not put tags like greyscale, monochrome in negative unless you understand what are you doing. Extra tags for brightness/colors/contrast section below can be usedArtist styles:Grids with examples https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg, list/wildcard https://files.catbox.moe/anleso.txt (also can be found in \"training data\").Used with \"by \" it's mandatory. It will not work properly without it.\"by \" is a meta-token for styles to avoid mixing/misinterpret with tags/characters of similar or close name. This allows to have a better results for styles and at the same time avoid random style fluctuation that you may observe in other checkpoints.Multiple give very interesting results, can be controlled with prompt weights and spells.YOU MUST ADD BREAK after artists/style tags (for A1111) or concat conditioning (for Comfy) or put them in the very end of your prompt.For example:by kantoku, by wlop, best quality, masterpiece BREAK 1girl, ...General styles:2.5d, anime screencap, bold line, sketch, cgi, digital painting, flat colors, smooth shading, minimalistic, ink style, oil style, pastel styleBooru tags styles:1950s (style), 1960s (style), 1970s (style), 1980s (style), 1990s (style), 2000s (style), animification, art nouveau, pinup (style), toon (style), western comics (style), nihonga, shikishi, minimalism, fine art parodyand everything from this group https://danbooru.donmai.us/wiki_pages/traditional_media.Can be used in combinations (with artists too), with weights, both in positive and negative prompts.Characters:Use full name booru tag and proper formatting, like karin_(blue_archive) -> karin \\(blue archive\\), use skin tags for better reproducing, like karin \\(bunny\\) \\(blue archive\\). Autocomplete extension might be very useful.Most characters are recognized just by their booru tag, but it will be more accurate if you describe their basic traits. Here you can easily redress your waifu/husbendo just by the prompt without suffering from the typical leaks of basic features.Natural text:Use it in combination with booru tags, works great. Use only natural text after typing styles and quality tags. Use just booru tags and forget about it, it's all up to you. To get best performance keep track if CLIP 75 tokens chunks.About 4M of images in dataset had hybrid natural-text captions, made by Claude, GPT, Gemini, ToriiGate https://huggingface.co/Minthy/ToriiGate-v0.4-7B, then refactored, cleaned and combined with tags in different variations for augmentation.Unlike typical captions, these contains character names which is very useful. Better to keep it clean, short and convenient description works best. Better not use long and sloppy BS likeA mysteriously enchanting feminine entity of indeterminate yet youthful essence, whose celestial visage radiates with the ethereal luminescence of a thousand dying stars, blessed with locks cascading like the golden rivers of ancient mythology, perhaps styled in a manner reminiscent of contemporary fashion trends though not necessarily adhering to any specific aesthetic paradigm. Her eyes, pools of unfathomable depth and hue, sparkle with the wisdom of millennia yet maintain an innocent quality that defies temporal constraints...For captioning you can use ToriiGate https://huggingface.co/Minthy/ToriiGate-v0.4-7B in short mode.And don't expect it to be as good as flux and others, it tries very hard and after several rolls usually you can get what you want, but it is not that stable and detailed.Lots of Tail/Ears-related concepts:Oh yeahtail censor, holding own tail, hugging own tail, holding another's tail, tail grab, tail raised, tail down, ears down, hand on own ear, tail around own leg, tail around penis, tailjob, tail through clothes, tail under clothes, lifted by tail, tail biting, tail penetration (including a specific indication of vaginal/anal), tail masturbation, holding with tail, panties on tail, bra on tail, tail focus, presenting own tail...(booru meaning, not e621) and many others with natural text. The majority works perfectly, some requires a lot of rolling.Brightness/colors/contrast:You can use extra meta tags to control it:low brightness, high brightness, low saturation, high saturation, low gamma, high gamma, sharp colors, soft colors, hdr, sdrExampleThey work both in epsilon and vpred version and works really good.Epsilon version relies on them too much. Without low brightness or low gamma or limited range (in negative) it might be difficult to achieve true 0,0,0 black, the same often true for white.Both epsilon and vpred versions have like true zsnr, full range of colors and brightness without common flaws observed. But they behaves differently, just try it.Vpred versionMain thing you need to know - lower your CFG from 7 down to 5 (or less). Otherwise, the use is similar with advantages.It seems that starting from v0.7 vpred works flawlessly now. It shouldn't suffer from ignorance of tags close to the 75tokens chunk borders like nai. It is more difficult to get burned images - even on cfg7 usually it just over-saturated but with smooth gradients, which can be useful for some styles. Yes it can make anything from (0,0,0) to (255,255,255). You will find brightness meta tags described above quite useful for easier/lazy prompting, natural text expressions also work. To get the most dark image - put high brightness into negative and/or use low brightness, low gamma tags. If you don't like very bright skin on dark background and want to reduce contrast (or on the contrary, enhance the effect) - use hdr/sdr in negative/positive.It was reported that in rare cases on some prompts there is a drop in contrast. Looks like other vpred models have same behaviour with such prompts, adding a \"separator\" closer to the border of the 75-token chunk fixes this. However, with 0.7 I haven't encountered this myself.To launch vpred version you will need dev build of A1111, Comfy (with special loader node), Forge or Reforge. Just use same parameters (Euler a, cfg 3..5, 20..28 steps) like epsilon. No need to use Cfg rescale, but you can try it, cfg++ works great.Base model:The model here has a small unet polishint after main training to improve small details, bump up resolution and others. Hovewer, you may be also interested into a RouWei-Base, which sometimes can perform better at complex prompts despite having minor mistakes in small details. It also comes in FP32, for example if you want to use fp32 text encoder nodes in Comfy, merge it or finetune.It can be found in Huggingface repo https://huggingface.co/Minthy/RouWei-0.8Known issues:Off course there are:Artists and style tags must be seperated into a different chunk from main prompt or come very lastThere may be some positional or combinational bias in rare cases, but it's not yet clear.There are some complaints about few of the general styles.Epsilon version relies too much on brightness meta tags, sometimes you will need to use them to get desired brightness shiftSome newly added styles/characters might be not as good and disctinct as they deserve toTo be discoveredRequests for artists/characters in future models are open. If you find artist/character/concept that perform weak, inaccurate or has strong watermark - please report, will add them explicitly. Follow for a new versions.JOIN THE DISCORD SERVER https://discord.gg/ZXHENAhqE9License:Same as illustrious. Fell free to use in your merges, finetunes, ets. but please leave a link or mention, it is mandatoryHow it's madeI'll consider to make a report or something like it later. For sure.In short, 98% of work is related to dataset preparations. Instead of blindly relying on loss-weighting based on tag frequency from nai paper, a custom guided loss-weighting implementation along with asynchronous collator for balancing have been used. Ztsnr (or close to it) with Epsilon prediction was achieved using noise scheduler augmentation.Spent compute - over 8k hours of H100 (apart from research and fail attempts)Thanks:First of all I'd like to acknowledge everyone who supports open source, develops in improves code. Thanks to the authors of illustrious for releasing model, thank to NoobAI team for being pioneers in open finetuning of such a scale, sharing experience, raising and solving issues that previously went unnoticed.Personal:Artists wish to remain anonymous for sharing private works; Few anonymous persons - donations, code, captions, etc., Soviet Cat - GPU sponsoring; Sv1. - llm access, captioning, code; K. - training code; Bakariso - datasets, testing, advices, insides; NeuroSenko - donations, testing, code; LOL2024 - a lot of unique datasets; T.,[] - datasets, testing, advises; rred, dga, Fi., ello - donations; TekeshiX - datasets. And other fellow brothers that helped. Love you so much \u2764\ufe0f.And off course everyone who made feedback and requests, it's really valuable.If I forgot to mention anyone, please notify.DonationsIf you want to support - share my models, leave feedback, make a cute picture with kemonomimi-girl. And of course, support original artists.AI is my hobby, I'm spending money on it and not begging for donations. However, it has turned into a large-scale and expensive undertaking. Consider to support to accelerate new training and researches.(Just keep in mind that I can waste it on alcohol or cosplay girls)BTC: bc1qwv83ggq8rvv07uk6dv4njs0j3yygj3aax4wg6cETH/USDT(e): 0x04C8a749F49aE8a56CB84cF0C99CD9E92eDB17dbXMR: 47F7JAyKP8tMBtzwxpoZsUVB8wzg2VrbtDKBice9FAS1FikbHEXXPof4PAb42CQ5ch8p8Hs4RvJuzPHDtaVSdQzD6ZbA5TZif you can offer gpu-time (a100+) - PM.",
     "sd version": "Other"
 }

Stable-diffusion/rouwei_v080Vpred.html

@@ -0,0 +1,250 @@
+
+                <head>
+                    
+                <meta charset="UTF-8">
+                    <link rel="stylesheet" type="text/css" href="/content/A1111/extensions/sd-civitai-browser-plus/style_html.css">
+                </head>
+                <div class="model-block">
+                    <h2><a href=https://civitai.com/models/950531 target="_blank" id="model_header">RouWei</a></h2>
+                    <h3 class="model-uploader">Uploaded by <a href="https://civitai.com/user/Minthybasis" target="_blank">Minthybasis</a><div class="avatar"><img src=https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/c3116f85-2b3b-4daf-a138-aed342f0cf2d/width=96/Minthybasis.jpeg></div></h3>
+                    <div class="civitai-version-info" style="display:flex; flex-wrap:wrap; justify-content:space-between;">
+                        <dl id="info_block">
+                            <dt>Version</dt>
+                            <dd>0.6 vpred</dd>
+                            <dt>Base Model</dt>
+                            <dd>Illustrious</dd>
+                            <dt>Published</dt>
+                            <dd>2024-11-15</dd>
+                            <dt>Availability</dt>
+                            <dd>Public</dd>
+                            <dt>CivitAI Tags</dt>
+                            <dd>
+                                <div class="civitai-tags-container">
+                                    <span class="civitai-tag">anime</span><span class="civitai-tag">base model</span>
+                                </div>
+                            </dd>
+                            <dt>Download Link</dt>
+                            <dd><a href=https://civitai.com/api/download/models/1064242 target="_blank">https://civitai.com/api/download/models/1064242</a></dd>
+                        </dl>
+                        <div style="align-self:center; min-width:320px;">
+                            <div>
+                                <p style="line-height: 2; font-weight: bold;"><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Use the model without crediting the creator<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Sell images they generate<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Run on services that generate images for money<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Run on Civitai<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Share merges using this model<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="red"><path d="M18 6l-12 12"></path><path d="M6 6l12 12"></path></svg></span> Sell this model or merges using this model<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="red"><path d="M18 6l-12 12"></path><path d="M6 6l12 12"></path></svg></span> Have different permissions when sharing merges</p>
+                            </div>
+                        </div>
+                    </div>
+                    <input type="checkbox" id="preview-civitai-description" class="description-toggle-checkbox">
+                    <div class="model-description">
+                        <h2>Description</h2>
+                        <h2 id="in-depth-retraining-of-illustrious-to-achieve-best-prompt-adherence-knowledge-and-state-of-the-art-performance.-66mudratg">In depth retraining of Illustrious to achieve best prompt adherence, knowledge and state of the art performance.</h2><p><span style="color:rgb(134, 142, 150)">Big dreams come true</span></p><p><span style="color:rgb(134, 142, 150)">The version number is just an index of current final release, not a fraction of the planned training.</span></p><p><a target="_blank" rel="ugc" href="https://huggingface.co/Minthy/RouWei-0.8"><span style="color:rgb(134, 142, 150)">HF repo</span></a></p><p></p><h3 id="large-scale-finetune-using-gpu-cluster-with-a-dataset-of-~13m-pictures-(~4m-with-natural-text-captions)-pqrwpp4f0">Large scale finetune using gpu cluster with a dataset of ~13M pictures (~4M with natural text captions)</h3><ul><li><p>Fresh and wast knowledge about characters, concepts, styles, cultural and related things</p></li><li><p>The best prompt adherence among SDXL anime models at the moment of release</p></li><li><p>Solved main problems with tags bleeding and biases, common for Illustrious, NoobAi and other checkpoints</p></li><li><p>Excellent aesthetics and knowledge across a wide range of styles (<a target="_blank" rel="ugc" href="https://files.catbox.moe/anleso.txt">over 50,000 artists</a> (<a rel="ugc" href="https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg">examples</a>), including hundreds of unique cherry-picked datasets from private galleries, including those received from the artists themselves)</p></li><li><p>High flexibility and variety without stability tradeoff</p></li><li><p>No more annoying watermarks for popular styles thanks to clean dataset</p></li><li><p>Vibrant colors and smooth gradients without trace of burning, full range even with epsilon</p></li><li><p>Pure training from Illustrious v0.1 without involving third-party checkpoints, Loras, tweakers, etc.</p></li></ul><p></p><h3 id="there-are-also-some-issues-and-changes-compared-to-the-previous-version-please-rtfm.-aaa28dbe6">There are also some issues and changes compared to the previous version, please RTFM.</h3><p></p><p>Dataset cut-off - end of April 2025.</p><p></p><h1 id="features-and-prompting:-j89sanl8e">Features and prompting:</h1><p></p><h2 id="important-change:-f6w18r0i0">Important change:</h2><p>When you are prompting artist styles, especially mixing several, <strong><u>their tags MUST BE in a separate CLIP chunk</u></strong>. Just add BREAK after it (for A1111 and derivatives), use conditioning concat node (for Comfy) <strong><u>or at least put them in the very end</u></strong>. Otherwise, significant degradation of results is likely.</p><p></p><h2 id="basic:-aroav4616">Basic:</h2><p></p><p>The checkpoint works both with short-simple and long-complex prompts. However, if there are contradictory or weird things - unlike with others they won't be ignored affecting the output. No guide-rails, no safeguards, no lobotomy.</p><p>Just prompt what you want to see and don't prompt what shouldn't be on the picture. If you want to have a view from above - don't put ceiling into positive, if you want to have crop view with head out of frame - don't make detailed description of character facial features, and so on. Pretty simple but sometimes people are missing it.</p><p></p><p>Version 0.8 comes with advanced understanding of natural text prompts. <u>It doesn't mean that you are obligated to use it</u>, tags only - completely fine, especially because understanding of tags combinations is also improved.</p><p>Do not expect it to perform like Flux or other models based on T5 or LLM text encoders. The whole size ot SDXL checkpoint is less then only that text encoder, in addition illustrious-v0.1 which is used as the base completely forgot a lot of general things from vanilla sdxl-base.</p><p>However, even in current state it works much better, allows to do new things usually impossible without external guidance, as well making manual editing, inpainting, etc more convenient.</p><p></p><p>To achieve best performance you should keep track of CLIP chunks. In SDXL the prompt is separated into a chunks of 75 (77 including BOS and EOS) tokens, that are processing by CLIP separately, and only then are concatinating and comes as conditions to unet.</p><p>If you want to specify some features for character/object and separate them from other prompt parts - make sure they are in the same chunk and optionally separate it with <code style="text-wrap: wrap">BREAK</code>. It will not solve problem of traits mixing completely, but can reduce it improving overall understanding, since text encoders on RouWei are able to process the whole sequence, not individual concepts better then others.</p><p>Dataset contains only booru-style tags and natural text expressions. Despite having a share of furries, real life photos, western media, etc. all captions have been converted to classic booru style to avoid a number of problems from mixing of different systems. So e621 tags won't be understanded properly.</p><p></p><h3 id="sampling-parameters:-7andwk13p">Sampling parameters:</h3><ul><li><p>~1 megapixel for txt2img, any AR with resolution multiple of 32 (1024x1024, 1056x, 1152x, 1216x832,...). Euler_a, 20..28steps.</p></li><li><p>CFG: for epsilon version 4..9 (7 is best), for vpred version, 3..5</p></li><li><p>Sigmas multiply may improve results a bit, CFG++ samplers work fine. LCM/PCM/DMD/... and exotic samplers untested.</p></li><li><p><strong>Some schedulers doesn't work well.</strong></p></li><li><p>Highresfix - x1.5 latent + denoise 0.6 or any gan + denoise 0.3..0.55.</p></li><li><p>For vpred version lower CFG 3..5 is needed!</p></li></ul><p></p><h3 id="for-vpred-version-lower-cfg-3..5-is-needed!-uwz9s21js"><strong>For vpred version lower CFG 3..5 is needed!</strong></h3><p></p><h2 id="quality-classification:-kw87tas72">Quality classification:</h2><p>Only 4 quality tags:</p><pre><code style="text-wrap: wrap">masterpiece, best quality</code></pre><p>for positive and</p><pre><code style="text-wrap: wrap">low quality, worst quality</code></pre><p>for negative.</p><p>Nothing else. Actually you can even omit positive and reduce negative to <code style="text-wrap: wrap">low quality</code> only, since they can affect basic style and composition.</p><p>Meta tags like <code style="text-wrap: wrap">lowres</code> have been removed and don't work, better not to use them. Low resolution images have been either removed or upscaled and cleaned with DAT depending on their importance.</p><p></p><h2 id="negative-prompt:-79eot1hhx">Negative prompt:</h2><pre><code style="text-wrap: wrap">worst quality, low quality, watermark</code></pre><p>That's all, no need of "rusty trombone", "farting on prey" and others. Do not put tags like greyscale, monochrome in negative unless you understand what are you doing. Extra tags for brightness/colors/contrast section below can be used</p><p></p><h2 id="artist-styles:-6yumq5hap">Artist styles:</h2><p></p><p><a target="_blank" rel="ugc" href="https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg">Grids with examples</a>, <a target="_blank" rel="ugc" href="https://files.catbox.moe/anleso.txt">list/wildcard</a> (also can be found in "training data").</p><p></p><h3 id="used-with-&quot;by-&quot;-it's-mandatory.-it-will-not-work-properly-without-it.-innc8cnap"><u>Used with "by " it's mandatory</u>. It will not work properly without it.</h3><p>"by " is a meta-token for styles to avoid mixing/misinterpret with tags/characters of similar or close name. This allows to have a better results for styles and at the same time avoid random style fluctuation that you may observe in other checkpoints.</p><p></p><p>Multiple give very interesting results, can be controlled with prompt weights and spells.</p><p></p><h3 id="you-must-add-break-after-artistsstyle-tags-(for-a1111)-or-concat-conditioning-(for-comfy)-or-put-them-in-the-very-end-of-your-prompt.-36i67w9c0">YOU MUST ADD <code style="text-wrap: wrap">BREAK</code> after artists/style tags (for A1111) or concat conditioning (for Comfy) or put them in the very end of your prompt.</h3><p>For example:</p><pre><code style="text-wrap: wrap">by kantoku, by wlop, best quality, masterpiece BREAK 1girl, ...</code></pre><p></p><h2 id="general-styles:-5ig9sgqza">General styles:</h2><pre><code style="text-wrap: wrap">2.5d, anime screencap, bold line, sketch, cgi, digital painting, flat colors, smooth shading, minimalistic, ink style, oil style, pastel style</code></pre><h3 id="booru-tags-styles:-uq05a7u5y">Booru tags styles:</h3><pre><code style="text-wrap: wrap">1950s (style), 1960s (style), 1970s (style), 1980s (style), 1990s (style), 2000s (style), animification, art nouveau, pinup (style), toon (style), western comics (style), nihonga, shikishi, minimalism, fine art parody</code></pre><p>and everything from <a target="_blank" rel="ugc" href="https://danbooru.donmai.us/wiki_pages/traditional_media">this group</a>.</p><p>Can be used in combinations (with artists too), with weights, both in positive and negative prompts.</p><p></p><h2 id="characters:-8htyrng9m">Characters:</h2><p>Use full name booru tag and proper formatting, like <code style="text-wrap: wrap">karin_(blue_archive)</code> -&gt; <code style="text-wrap: wrap">karin \(blue archive\)</code>, use skin tags for better reproducing, like <code style="text-wrap: wrap">karin \(bunny\) \(blue archive\)</code>. Autocomplete extension might be very useful.</p><p>Most characters are recognized just by their booru tag, but it will be more accurate if you describe their basic traits. Here you can easily redress your waifu/husbendo just by the prompt without suffering from the typical leaks of basic features.</p><p></p><h2 id="natural-text:-cq8whprlj">Natural text:</h2><p>Use it in combination with booru tags, works great. Use only natural text after typing styles and quality tags. Use just booru tags and forget about it, it's all up to you. To get best performance keep track if CLIP 75 tokens chunks.</p><p></p><p>About 4M of images in dataset had hybrid natural-text captions, made by Claude, GPT, Gemini, <a target="_blank" rel="ugc" href="https://huggingface.co/Minthy/ToriiGate-v0.4-7B">ToriiGate</a>, then refactored, cleaned and combined with tags in different variations for augmentation.</p><p></p><p>Unlike typical captions, these contains character names which is very useful. Better to keep it clean, short and convenient description works best. Better <strong><u>not</u></strong> use long and sloppy BS like</p><pre><code style="text-wrap: wrap">A mysteriously enchanting feminine entity of indeterminate yet youthful essence, whose celestial visage radiates with the ethereal luminescence of a thousand dying stars, blessed with locks cascading like the golden rivers of ancient mythology, perhaps styled in a manner reminiscent of contemporary fashion trends though not necessarily adhering to any specific aesthetic paradigm. Her eyes, pools of unfathomable depth and hue, sparkle with the wisdom of millennia yet maintain an innocent quality that defies temporal constraints...</code></pre><p>For captioning you can use <a target="_blank" rel="ugc" href="https://huggingface.co/Minthy/ToriiGate-v0.4-7B">ToriiGate</a> in short mode.</p><p></p><p>And don't expect it to be as good as flux and others, it tries very hard and after several rolls usually you can get what you want, but it is not that stable and detailed.</p><p></p><h2 id="lots-of-tailears-related-concepts:-qsxmk9gzm">Lots of Tail/Ears-related concepts:</h2><p>Oh yeah</p><pre><code style="text-wrap: wrap">tail censor, holding own tail, hugging own tail, holding another's tail, tail grab, tail raised, tail down, ears down, hand on own ear, tail around own leg, tail around penis, tailjob, tail through clothes, tail under clothes, lifted by tail, tail biting, tail penetration (including a specific indication of vaginal/anal), tail masturbation, holding with tail, panties on tail, bra on tail, tail focus, presenting own tail...</code></pre><p>(booru meaning, not e621) and many others with natural text. The majority works perfectly, some requires a lot of rolling.</p><p></p><h1 id="brightnesscolorscontrast:-lk2g7nwtj">Brightness/colors/contrast:</h1><p>You can use extra meta tags to control it:</p><pre><code style="text-wrap: wrap">low brightness, high brightness, low saturation, high saturation, low gamma, high gamma, sharp colors, soft colors, hdr, sdr</code></pre><p><a target="_blank" rel="ugc" href="https://files.catbox.moe/r8wrb6.jpg">Example</a></p><p>They work both in epsilon and vpred version and works really good.</p><p>Epsilon version relies on them too much. Without low brightness or low gamma or limited range (in negative) it might be difficult to achieve true 0,0,0 black, the same often true for white.</p><p>Both epsilon and vpred versions have like true zsnr, full range of colors and brightness without common flaws observed. But they behaves differently, just try it.</p><p></p><h1 id="vpred-version-9pr2w5h4d">Vpred version</h1><p></p><p>Main thing you need to know - lower your CFG from 7 down to 5 (or less). Otherwise, the use is similar with advantages.</p><p>It seems that starting from v0.7 vpred works flawlessly now. It shouldn't suffer from ignorance of tags close to the 75tokens chunk borders like nai. It is more difficult to get burned images - even on cfg7 usually it just over-saturated but with smooth gradients, which can be useful for some styles. Yes it can make anything from (0,0,0) to (255,255,255). You will find brightness meta tags described above quite useful for easier/lazy prompting, natural text expressions also work. To get the most dark image - put <code style="text-wrap: wrap">high brightness</code> into <u>negative</u> and/or use <code style="text-wrap: wrap">low brightness, low gamma</code> tags. If you don't like very bright skin on dark background and want to reduce contrast (or on the contrary, enhance the effect) - use hdr/sdr in negative/positive.</p><p>It was reported that in rare cases on some prompts there is a drop in contrast. Looks like other vpred models have same behaviour with such prompts, adding a "separator" closer to the border of the 75-token chunk fixes this. However, with 0.7 I haven't encountered this myself.</p><p></p><p><span style="color:rgb(134, 142, 150)">To launch vpred version you will need dev build of A1111, Comfy (with special loader node), Forge or Reforge. Just use same parameters (Euler a, cfg 3..5, 20..28 steps) like epsilon. No need to use Cfg rescale, but you can try it, cfg++ works great.</span></p><p></p><h1 id="base-model:-lwa0dfdbf">Base model:</h1><p>The model here has a small unet polishint after main training to improve small details, bump up resolution and others. Hovewer, you may be also interested into a RouWei-Base, which sometimes can perform better at complex prompts despite having minor mistakes in small details. It also comes in FP32, for example if you want to use fp32 text encoder nodes in Comfy, merge it or finetune.</p><p>It can be found in <a target="_blank" rel="ugc" href="https://huggingface.co/Minthy/RouWei-0.8">Huggingface repo</a></p><p></p><h1 id="known-issues:-jxgt6yemu">Known issues:</h1><p>Off course there are:</p><ul><li><p>Artists and style tags must be seperated into a different chunk from main prompt or come very last</p></li><li><p>There may be some positional or combinational bias in rare cases, but it's not yet clear.</p></li><li><p>There are some complaints about few of the general styles.</p></li><li><p>Epsilon version relies too much on brightness meta tags, sometimes you will need to use them to get desired brightness shift</p></li><li><p>Some newly added styles/characters might be not as good and disctinct as they deserve to</p></li><li><p>To be discovered</p><p></p></li><li><p></p></li></ul><h3 id="requests-for-artistscharacters-in-future-models-are-open.-if-you-find-artistcharacterconcept-that-perform-weak-inaccurate-or-has-strong-watermark-please-report-will-add-them-explicitly.-follow-for-a-new-versions.-kgc39ed4e">Requests for artists/characters in future models are open. If you find artist/character/concept that perform weak, inaccurate or has strong watermark - please report, will add them explicitly. Follow for a new versions.</h3><p></p><h2 id="join-the-discord-server-9lpaojolv"><a target="_blank" rel="ugc" href="https://discord.gg/ZXHENAhqE9">JOIN THE DISCORD SERVER</a></h2><p></p><h2 id="license:-0hpy342uh">License:</h2><p>Same as illustrious. Fell free to use in your merges, finetunes, ets. but please leave a link or mention, it is mandatory</p><p></p><h2 id="how-it's-made-zc68i8qve">How it's made</h2><p>I'll consider to make a report or something like it later. For sure.</p><p>In short, 98% of work is related to dataset preparations. Instead of blindly relying on loss-weighting based on tag frequency from nai paper, a custom guided loss-weighting implementation along with asynchronous collator for balancing have been used. Ztsnr (or close to it) with Epsilon prediction was achieved using noise scheduler augmentation.</p><p></p><p>Spent compute - over 8k hours of H100 (apart from research and fail attempts)</p><p></p><h1 id="thanks:-4tsy2pj9q">Thanks:</h1><p>First of all I'd like to acknowledge everyone who supports open source, develops in improves code. Thanks to the authors of illustrious for releasing model, thank to NoobAI team for being pioneers in open finetuning of such a scale, sharing experience, raising and solving issues that previously went unnoticed.</p><h3 id="personal:-9a91vqx2i">Personal:</h3><p>Artists wish to remain anonymous for sharing private works; Few anonymous persons - donations, code, captions, etc., Soviet Cat - GPU sponsoring; Sv1. - llm access, captioning, code; K. - training code; Bakariso - datasets, testing, advices, insides; NeuroSenko - donations, testing, code; LOL2024 - a lot of unique datasets; T.,[] - datasets, testing, advises; rred, dga, Fi., ello - donations; TekeshiX - datasets. And other fellow brothers that helped. Love you so much ❤️.</p><p>And off course everyone who made feedback and requests, it's really valuable.</p><p><span style="color:rgb(134, 142, 150)">If I forgot to mention anyone, please notify.</span></p><p></p><h2 id="donations-jdo27wd0z">Donations</h2><p>If you want to support - share my models, leave feedback, make a cute picture with kemonomimi-girl. And of course, support original artists.</p><p>AI is my hobby, I'm spending money on it and not begging for donations. However, it has turned into a large-scale and expensive undertaking. Consider to support to accelerate new training and researches.</p><p><span style="color:rgb(134, 142, 150)">(Just keep in mind that I can waste it on alcohol or cosplay girls)</span></p><p>BTC: bc1qwv83ggq8rvv07uk6dv4njs0j3yygj3aax4wg6c</p><p>ETH/USDT(e): 0x04C8a749F49aE8a56CB84cF0C99CD9E92eDB17db</p><p>XMR: 47F7JAyKP8tMBtzwxpoZsUVB8wzg2VrbtDKBice9FAS1FikbHEXXPof4PAb42CQ5ch8p8Hs4RvJuzPHDtaVSdQzD6ZbA5TZ</p><p>if you can offer gpu-time (a100+) - PM.</p><p></p><p></p><p></p><p></p><p></p>
+                    </div>
+                    <label for="preview-civitai-description" class="description-toggle-label"></label>
+                </div>
+                <div align=center><div class="sampleimgs"><input type="radio" name="zoomRadio" id="resetZoom" class="zoom-radio" checked>
+                    <div class="model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_0" class="zoom-radio">
+                    <label for="zoomRadiopreview_0" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/f96c7ce9-0eaf-4461-bee3-176a008ee8d2/width=1536/40258578.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/f96c7ce9-0eaf-4461-bee3-176a008ee8d2/width=1536/40258578.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>masterpiece, best quality, good quality,
+1girl, konpaku youmu, half-closed eyes, solo, black hairband, (konpaku youmu (ghost):1.2), hairband, green vest, puffy sleeves, upper body, bow, black bowtie, black bow, short hair, vest, bowtie, ribbon, leaf, shirt, wind, short sleeves, black background, looking down, white shirt, hitodama, black ribbon, hair ribbon, puffy short sleeves, grey hair, simple background, floating hair, closed mouth, expressionless, wing collar, hitodama print, buttons, green eyes, collared shirt, falling leaves, aqua eyes, glowing eyes, pale skin, (skeleton, split screen, straight-on, android, crank, gauge, gears, glowing, light bulb, mechabare, sparkle, x-ray:1.2), 
+&lt;lora:youmuanon_noob-vpred0.1_flawless_ds_v2:1&gt;</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, bad quality, lowres, jpeg artifacts, bad anatomy, bad hands, multiple views, signature, watermark, censored, covered navel, limited range, high brightness, high contrast,</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>4147779862</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1024x1024</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred_rc1</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>5</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_0">
+                                    <label class="tab-label" for="chckpreview_0">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl-vae-fp16-fix.safetensors</dd></div><div class="civitai-meta"><dt>Ensd</dt><dd>31337</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;62c7c729ad&#x27;, &#x27;model&#x27;: &#x27;e63d568c63&#x27;, &#x27;lora:youmuanon_noob-vpred0.1_flawless_ds_v2&#x27;: &#x27;7b7e860a9bae&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.3.6dev2-v1.10.1RC-latest-1453-g0fd79513</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;7b7e860a9bae&#x27;, &#x27;name&#x27;: &#x27;youmuanon_noob-vpred0.1_flawless_ds_v2&#x27;, &#x27;type&#x27;: &#x27;lora&#x27;, &#x27;weight&#x27;: 1}, {&#x27;hash&#x27;: &#x27;e63d568c63&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred_rc1&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>e63d568c63</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Align Your Steps</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>2x-AnimeSharpV3</dd></div><div class="civitai-meta"><dt>Adetailer model</dt><dd>face_yolov8n.pt</dd></div><div class="civitai-meta"><dt>Adetailer version</dt><dd>24.9.0</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Sigmas_merge_mode</dt><dd>Multiply</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.5</dd></div><div class="civitai-meta"><dt>Sigmas_mult_factor</dt><dd>2</dd></div><div class="civitai-meta"><dt>Adetailer mask blur</dt><dd>8</dd></div><div class="civitai-meta"><dt>Adetailer confidence</dt><dd>0.3</dd></div><div class="civitai-meta"><dt>Sigmas_merge_enabled</dt><dd>True</dd></div><div class="civitai-meta"><dt>Adetailer dilate erode</dt><dd>4</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Adetailer inpaint padding</dt><dd>32</dd></div><div class="civitai-meta"><dt>Adetailer negative prompt</dt><dd>{}</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div><div class="civitai-meta"><dt>Adetailer denoising strength</dt><dd>0.4</dd></div><div class="civitai-meta"><dt>Old prompt editing timelines</dt><dd>True</dd></div><div class="civitai-meta"><dt>Adetailer inpaint only masked</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_1" class="zoom-radio">
+                    <label for="zoomRadiopreview_1" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/76198546-76cd-4bf4-9fc8-f935da633bbf/width=1328/40258585.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/76198546-76cd-4bf4-9fc8-f935da633bbf/width=1328/40258585.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>(by deel \(rkeg\):0.8), best quality, masterpiece,
+1girl, folinic \(arknights\), cat girl, brown hair, multicolored hair, cat ears, surprised, blush, hair flower, holding papers, BREAK
+indoors, scientific lab, glass floor, intricate interior, computers, research facility</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, watermark</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>2290774044</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>832x1216</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_1">
+                                    <label class="tab-label" for="chckpreview_1">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;84c8888446&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;84c8888446&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>84c8888446</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.6</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_2" class="zoom-radio">
+                    <label for="zoomRadiopreview_2" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/5033de43-81ec-418e-9429-f2a7425353c4/width=1728/40258592.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/5033de43-81ec-418e-9429-f2a7425353c4/width=1728/40258592.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by ratatatat74, by tarakanovich,cube85, masterpiece, best quality,
+1girl, fake animal ears, cat ears, long black hair, large breasts, erect nipples, completely nude, butt plug, black thighhighs, garter straps, squatting, reverse cowgirl position, girl on top, from behind, ass, mature female, pussy juice, ass grab, large penis, vaginal, pov hands, tally, (slap mark:0.7),
+indoors, bedroom, dark night, dimly lit room, dark blue theme, low brightness, low gamma, sdr</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, limited range</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>1178610355</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1152x896</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_2">
+                                    <label class="tab-label" for="chckpreview_2">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_3" class="zoom-radio">
+                    <label for="zoomRadiopreview_3" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/9af10cd7-4513-4147-a4f2-033f46afd3e0/width=1824/40258577.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/9af10cd7-4513-4147-a4f2-033f46afd3e0/width=1824/40258577.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>(by ratatatat74:0.7), by tarakanovich, cube85, masterpiece, best quality,
+1girl, vulpisfoglia \(arknights\), fox girl, fox ears, large ears, ears down, blonde hair, large breasts, yellow eyes, fox tail, nightgown, see-through, bottomless, from behind, tail censor, (scar on animal ear:0.8), white panties, naughty face,
+all fours, fox shadow puppet, glowing eyes,
+black background, low brightness, low gamma</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, extra ears</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>1556719420</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1216x832</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>20</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_3">
+                                    <label class="tab-label" for="chckpreview_3">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_4" class="zoom-radio">
+                    <label for="zoomRadiopreview_4" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/0d48a8dd-f3d3-42f6-b79d-de7de7b52139/width=1200/40258590.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/0d48a8dd-f3d3-42f6-b79d-de7de7b52139/width=1200/40258590.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>masterpiece, best quality, by merrytail, by sub-res, by kie \(wylee2212\),
+1girl, fox devil \(chainsaw man\), fox girl, fox ears, fox tail, whisker markings, long white hair, red eyes, (extra eyes:1.3), fangs, evil smile, white kimono with floral print, collarbone, hair bun, fox shadow puppet, cowboy shot, posing,
+black background, low brightness</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, limited range, monster, 1other, animal, fox</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>1376750140</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>800x1344</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_4">
+                                    <label class="tab-label" for="chckpreview_4">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_5" class="zoom-radio">
+                    <label for="zoomRadiopreview_5" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/209ff688-e01d-41d4-b842-a3ffadd3741f/width=1824/40258575.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/209ff688-e01d-41d4-b842-a3ffadd3741f/width=1824/40258575.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>masterpiece, best quality, by done \(donezumi\),
+1girl, kon \(tokyo ravens\), tokyo ravens, fox girl, fox ears, fox tail, white hair, thick eyebrows, pale skin, blue eyes, white aqua japanese clothes, detached sleeves, ribbon trim, fox shadow puppet, pom pom \(clothes\), black shorts, white thighhighs, hips, happy, BREAK floating in air, legs apart, posing, full body, wide shot,
+white background, high brightness</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>2168859150</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1216x832</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>20</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_5">
+                                    <label class="tab-label" for="chckpreview_5">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;84c8888446&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;84c8888446&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>84c8888446</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_6" class="zoom-radio">
+                    <label for="zoomRadiopreview_6" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/90435b3a-ef7d-4be2-b700-00601088c553/width=1248/40258594.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/90435b3a-ef7d-4be2-b700-00601088c553/width=1248/40258594.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by tandohark, (by personal ami:0.6), best quality, masterpiece,
+2girls, asymmetrical docking BREAK
+izuna \(blue archive\), fox girl, fox ears, fox mask, fox tail, halo, topless, denim shorts, striped bikini bra, medium breasts, sideboob, visors cap BREAK
+wakamo \(blue archive\), fox girl, fox ears, fox tail, long black hair, red halo, hair ornament, large breasts, fox mask, white bikini swimsuit, collarbone, :3 BREAK
+outdoors, beach, seashore, palms, clear blue sky</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, watermark</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>1114850771</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>832x1216</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>5</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_6">
+                                    <label class="tab-label" for="chckpreview_6">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;84c8888446&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;84c8888446&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>84c8888446</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_7" class="zoom-radio">
+                    <label for="zoomRadiopreview_7" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/23c627f5-451b-47de-a1d3-7dce53adb8d9/width=1248/40258589.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/23c627f5-451b-47de-a1d3-7dce53adb8d9/width=1248/40258589.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by akai sashimi, by nat the lich, by foxykuro, by nekoya \(liu\), masterpiece,  best quality,
+1girl, kal&#x27;tsit \(arknights\), cat girl, lynx ears, silver hair, green eyes/, naked towel, covered nipples, off shoulder, oripathy lesion \(arknights\), collarbone, shaded face, (evil smile:1.1), smug, open mouth, angry, fang, tilted head, wet,
+holding handcuffs,
+cowboy shot, motion lines,
+indoors, office, dimmed, dark background</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, watermark, dutch angle</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>2542436241</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>832x1216</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_7">
+                                    <label class="tab-label" for="chckpreview_7">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Ensd</dt><dd>31337</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_8" class="zoom-radio">
+                    <label for="zoomRadiopreview_8" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/fd31b937-db26-404f-ba61-c86eecedb37c/width=1328/40258587.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/fd31b937-db26-404f-ba61-c86eecedb37c/width=1328/40258587.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by akai sashimi, by nat the lich, by foxykuro, by nekoya \(liu\), masterpiece, best quality,
+1girl, kal&#x27;tsit \(arknights\), cat girl, lynx ears, silver hair, green eyes, short tail, completely nude, oripathy lesion \(arknights\), scars on spine, reverse cowgirl position, motion lines, from below, open mouth, smug, heavenly ass, v,
+1boy, vaginal, sex, pov, deep penetration, pussy juice trail,
+indoors, bedroom, dimmed, dark background</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, watermark, abstract background</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>3956367149</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>832x1216</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>20</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_8">
+                                    <label class="tab-label" for="chckpreview_8">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.6</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_9" class="zoom-radio">
+                    <label for="zoomRadiopreview_9" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/11c49fc0-f6fa-40b2-bb86-30e3a2c9cd0a/width=2064/40258599.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/11c49fc0-f6fa-40b2-bb86-30e3a2c9cd0a/width=2064/40258599.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by kate fox, masterpiece, best quality,
+1girl, yuel \(granblue fantasy\), fox girl, yellow eyes, fox ears, large fox tail, bikini bra, fur trim, elbow gloves, :3, wide shot, BREAK
+outdoors, torii, asian nature, epic landscape of mountains, above clouds</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, bad hands, watermark
+worst quality, low quality</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>2135703813</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1376x720</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>20</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_9">
+                                    <label class="tab-label" for="chckpreview_9">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;84c8888446&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;84c8888446&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>84c8888446</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Rescale_cfg_enabled</dt><dd>True</dd></div><div class="civitai-meta"><dt>Rescale_cfg_multiplier</dt><dd>0.7</dd></div></div></div></div></dl></div></div></div></div>
+                

Stable-diffusion/rouwei_v080Vpred.json

@@ -0,0 +1,8 @@
+{
+    "sha256": "1A40B1BABCA6A143DE9E98BCC4EF1277FF688B33892DF5DB683EE25753C0FD3A",
+    "modelId": 950531,
+    "modelVersionId": 1882934,
+    "activation text": "",
+    "description": "In depth retraining of Illustrious to achieve best prompt adherence, knowledge and state of the art performance.Big dreams come trueThe version number is just an index of current final release, not a fraction of the planned training.HF repo https://huggingface.co/Minthy/RouWei-0.8Large scale finetune using gpu cluster with a dataset of ~13M pictures (~4M with natural text captions)Fresh and wast knowledge about characters, concepts, styles, cultural and related thingsThe best prompt adherence among SDXL anime models at the moment of releaseSolved main problems with tags bleeding and biases, common for Illustrious, NoobAi and other checkpointsExcellent aesthetics and knowledge across a wide range of styles (over 50,000 artists https://files.catbox.moe/anleso.txt (examples https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg), including hundreds of unique cherry-picked datasets from private galleries, including those received from the artists themselves)High flexibility and variety without stability tradeoffNo more annoying watermarks for popular styles thanks to clean datasetVibrant colors and smooth gradients without trace of burning, full range even with epsilonPure training from Illustrious v0.1 without involving third-party checkpoints, Loras, tweakers, etc.There are also some issues and changes compared to the previous version, please RTFM.Dataset cut-off - end of April 2025.Features and prompting:Important change:When you are prompting artist styles, especially mixing several, their tags MUST BE in a separate CLIP chunk. Just add BREAK after it (for A1111 and derivatives), use conditioning concat node (for Comfy) or at least put them in the very end. Otherwise, significant degradation of results is likely.Basic:The checkpoint works both with short-simple and long-complex prompts. However, if there are contradictory or weird things - unlike with others they won't be ignored affecting the output. No guide-rails, no safeguards, no lobotomy.Just prompt what you want to see and don't prompt what shouldn't be on the picture. If you want to have a view from above - don't put ceiling into positive, if you want to have crop view with head out of frame - don't make detailed description of character facial features, and so on. Pretty simple but sometimes people are missing it.Version 0.8 comes with advanced understanding of natural text prompts. It doesn't mean that you are obligated to use it, tags only - completely fine, especially because understanding of tags combinations is also improved.Do not expect it to perform like Flux or other models based on T5 or LLM text encoders. The whole size ot SDXL checkpoint is less then only that text encoder, in addition illustrious-v0.1 which is used as the base completely forgot a lot of general things from vanilla sdxl-base.However, even in current state it works much better, allows to do new things usually impossible without external guidance, as well making manual editing, inpainting, etc more convenient.To achieve best performance you should keep track of CLIP chunks. In SDXL the prompt is separated into a chunks of 75 (77 including BOS and EOS) tokens, that are processing by CLIP separately, and only then are concatinating and comes as conditions to unet.If you want to specify some features for character/object and separate them from other prompt parts - make sure they are in the same chunk and optionally separate it with BREAK. It will not solve problem of traits mixing completely, but can reduce it improving overall understanding, since text encoders on RouWei are able to process the whole sequence, not individual concepts better then others.Dataset contains only booru-style tags and natural text expressions. Despite having a share of furries, real life photos, western media, etc. all captions have been converted to classic booru style to avoid a number of problems from mixing of different systems. So e621 tags won't be understanded properly.Sampling parameters:~1 megapixel for txt2img, any AR with resolution multiple of 32 (1024x1024, 1056x, 1152x, 1216x832,...). Euler_a, 20..28steps.CFG: for epsilon version 4..9 (7 is best), for vpred version, 3..5Sigmas multiply may improve results a bit, CFG++ samplers work fine. LCM/PCM/DMD/... and exotic samplers untested.Some schedulers doesn't work well.Highresfix - x1.5 latent + denoise 0.6 or any gan + denoise 0.3..0.55.For vpred version lower CFG 3..5 is needed!For vpred version lower CFG 3..5 is needed!Quality classification:Only 4 quality tags:masterpiece, best qualityfor positive andlow quality, worst qualityfor negative.Nothing else. Actually you can even omit positive and reduce negative to low quality only, since they can affect basic style and composition.Meta tags like lowres have been removed and don't work, better not to use them. Low resolution images have been either removed or upscaled and cleaned with DAT depending on their importance.Negative prompt:worst quality, low quality, watermarkThat's all, no need of \"rusty trombone\", \"farting on prey\" and others. Do not put tags like greyscale, monochrome in negative unless you understand what are you doing. Extra tags for brightness/colors/contrast section below can be usedArtist styles:Grids with examples https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg, list/wildcard https://files.catbox.moe/anleso.txt (also can be found in \"training data\").Used with \"by \" it's mandatory. It will not work properly without it.\"by \" is a meta-token for styles to avoid mixing/misinterpret with tags/characters of similar or close name. This allows to have a better results for styles and at the same time avoid random style fluctuation that you may observe in other checkpoints.Multiple give very interesting results, can be controlled with prompt weights and spells.YOU MUST ADD BREAK after artists/style tags (for A1111) or concat conditioning (for Comfy) or put them in the very end of your prompt.For example:by kantoku, by wlop, best quality, masterpiece BREAK 1girl, ...General styles:2.5d, anime screencap, bold line, sketch, cgi, digital painting, flat colors, smooth shading, minimalistic, ink style, oil style, pastel styleBooru tags styles:1950s (style), 1960s (style), 1970s (style), 1980s (style), 1990s (style), 2000s (style), animification, art nouveau, pinup (style), toon (style), western comics (style), nihonga, shikishi, minimalism, fine art parodyand everything from this group https://danbooru.donmai.us/wiki_pages/traditional_media.Can be used in combinations (with artists too), with weights, both in positive and negative prompts.Characters:Use full name booru tag and proper formatting, like karin_(blue_archive) -> karin \\(blue archive\\), use skin tags for better reproducing, like karin \\(bunny\\) \\(blue archive\\). Autocomplete extension might be very useful.Most characters are recognized just by their booru tag, but it will be more accurate if you describe their basic traits. Here you can easily redress your waifu/husbendo just by the prompt without suffering from the typical leaks of basic features.Natural text:Use it in combination with booru tags, works great. Use only natural text after typing styles and quality tags. Use just booru tags and forget about it, it's all up to you. To get best performance keep track if CLIP 75 tokens chunks.About 4M of images in dataset had hybrid natural-text captions, made by Claude, GPT, Gemini, ToriiGate https://huggingface.co/Minthy/ToriiGate-v0.4-7B, then refactored, cleaned and combined with tags in different variations for augmentation.Unlike typical captions, these contains character names which is very useful. Better to keep it clean, short and convenient description works best. Better not use long and sloppy BS likeA mysteriously enchanting feminine entity of indeterminate yet youthful essence, whose celestial visage radiates with the ethereal luminescence of a thousand dying stars, blessed with locks cascading like the golden rivers of ancient mythology, perhaps styled in a manner reminiscent of contemporary fashion trends though not necessarily adhering to any specific aesthetic paradigm. Her eyes, pools of unfathomable depth and hue, sparkle with the wisdom of millennia yet maintain an innocent quality that defies temporal constraints...For captioning you can use ToriiGate https://huggingface.co/Minthy/ToriiGate-v0.4-7B in short mode.And don't expect it to be as good as flux and others, it tries very hard and after several rolls usually you can get what you want, but it is not that stable and detailed.Lots of Tail/Ears-related concepts:Oh yeahtail censor, holding own tail, hugging own tail, holding another's tail, tail grab, tail raised, tail down, ears down, hand on own ear, tail around own leg, tail around penis, tailjob, tail through clothes, tail under clothes, lifted by tail, tail biting, tail penetration (including a specific indication of vaginal/anal), tail masturbation, holding with tail, panties on tail, bra on tail, tail focus, presenting own tail...(booru meaning, not e621) and many others with natural text. The majority works perfectly, some requires a lot of rolling.Brightness/colors/contrast:You can use extra meta tags to control it:low brightness, high brightness, low saturation, high saturation, low gamma, high gamma, sharp colors, soft colors, hdr, sdrExampleThey work both in epsilon and vpred version and works really good.Epsilon version relies on them too much. Without low brightness or low gamma or limited range (in negative) it might be difficult to achieve true 0,0,0 black, the same often true for white.Both epsilon and vpred versions have like true zsnr, full range of colors and brightness without common flaws observed. But they behaves differently, just try it.Vpred versionMain thing you need to know - lower your CFG from 7 down to 5 (or less). Otherwise, the use is similar with advantages.It seems that starting from v0.7 vpred works flawlessly now. It shouldn't suffer from ignorance of tags close to the 75tokens chunk borders like nai. It is more difficult to get burned images - even on cfg7 usually it just over-saturated but with smooth gradients, which can be useful for some styles. Yes it can make anything from (0,0,0) to (255,255,255). You will find brightness meta tags described above quite useful for easier/lazy prompting, natural text expressions also work. To get the most dark image - put high brightness into negative and/or use low brightness, low gamma tags. If you don't like very bright skin on dark background and want to reduce contrast (or on the contrary, enhance the effect) - use hdr/sdr in negative/positive.It was reported that in rare cases on some prompts there is a drop in contrast. Looks like other vpred models have same behaviour with such prompts, adding a \"separator\" closer to the border of the 75-token chunk fixes this. However, with 0.7 I haven't encountered this myself.To launch vpred version you will need dev build of A1111, Comfy (with special loader node), Forge or Reforge. Just use same parameters (Euler a, cfg 3..5, 20..28 steps) like epsilon. No need to use Cfg rescale, but you can try it, cfg++ works great.Base model:The model here has a small unet polishint after main training to improve small details, bump up resolution and others. Hovewer, you may be also interested into a RouWei-Base, which sometimes can perform better at complex prompts despite having minor mistakes in small details. It also comes in FP32, for example if you want to use fp32 text encoder nodes in Comfy, merge it or finetune.It can be found in Huggingface repo https://huggingface.co/Minthy/RouWei-0.8Known issues:Off course there are:Artists and style tags must be seperated into a different chunk from main prompt or come very lastThere may be some positional or combinational bias in rare cases, but it's not yet clear.There are some complaints about few of the general styles.Epsilon version relies too much on brightness meta tags, sometimes you will need to use them to get desired brightness shiftSome newly added styles/characters might be not as good and disctinct as they deserve toTo be discoveredRequests for artists/characters in future models are open. If you find artist/character/concept that perform weak, inaccurate or has strong watermark - please report, will add them explicitly. Follow for a new versions.JOIN THE DISCORD SERVER https://discord.gg/ZXHENAhqE9License:Same as illustrious. Fell free to use in your merges, finetunes, ets. but please leave a link or mention, it is mandatoryHow it's madeI'll consider to make a report or something like it later. For sure.In short, 98% of work is related to dataset preparations. Instead of blindly relying on loss-weighting based on tag frequency from nai paper, a custom guided loss-weighting implementation along with asynchronous collator for balancing have been used. Ztsnr (or close to it) with Epsilon prediction was achieved using noise scheduler augmentation.Spent compute - over 8k hours of H100 (apart from research and fail attempts)Thanks:First of all I'd like to acknowledge everyone who supports open source, develops in improves code. Thanks to the authors of illustrious for releasing model, thank to NoobAI team for being pioneers in open finetuning of such a scale, sharing experience, raising and solving issues that previously went unnoticed.Personal:Artists wish to remain anonymous for sharing private works; Few anonymous persons - donations, code, captions, etc., Soviet Cat - GPU sponsoring; Sv1. - llm access, captioning, code; K. - training code; Bakariso - datasets, testing, advices, insides; NeuroSenko - donations, testing, code; LOL2024 - a lot of unique datasets; T.,[] - datasets, testing, advises; rred, dga, Fi., ello - donations; TekeshiX - datasets. And other fellow brothers that helped. Love you so much \u2764\ufe0f.And off course everyone who made feedback and requests, it's really valuable.If I forgot to mention anyone, please notify.DonationsIf you want to support - share my models, leave feedback, make a cute picture with kemonomimi-girl. And of course, support original artists.AI is my hobby, I'm spending money on it and not begging for donations. However, it has turned into a large-scale and expensive undertaking. Consider to support to accelerate new training and researches.(Just keep in mind that I can waste it on alcohol or cosplay girls)BTC: bc1qwv83ggq8rvv07uk6dv4njs0j3yygj3aax4wg6cETH/USDT(e): 0x04C8a749F49aE8a56CB84cF0C99CD9E92eDB17dbXMR: 47F7JAyKP8tMBtzwxpoZsUVB8wzg2VrbtDKBice9FAS1FikbHEXXPof4PAb42CQ5ch8p8Hs4RvJuzPHDtaVSdQzD6ZbA5TZif you can offer gpu-time (a100+) - PM.",
+    "sd version": "Other"
+}

Stable-diffusion/rouwei_v080Vpred.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a40b1babca6a143de9e98bcc4ef1277ff688b33892df5db683ee25753c0fd3a
+size 6938040512

Stable-diffusion/rouwei_v080Vpred_trainingData.html

@@ -0,0 +1,250 @@
+
+                <head>
+                    
+                <meta charset="UTF-8">
+                    <link rel="stylesheet" type="text/css" href="/content/A1111/extensions/sd-civitai-browser-plus/style_html.css">
+                </head>
+                <div class="model-block">
+                    <h2><a href=https://civitai.com/models/950531 target="_blank" id="model_header">RouWei</a></h2>
+                    <h3 class="model-uploader">Uploaded by <a href="https://civitai.com/user/Minthybasis" target="_blank">Minthybasis</a><div class="avatar"><img src=https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/c3116f85-2b3b-4daf-a138-aed342f0cf2d/width=96/Minthybasis.jpeg></div></h3>
+                    <div class="civitai-version-info" style="display:flex; flex-wrap:wrap; justify-content:space-between;">
+                        <dl id="info_block">
+                            <dt>Version</dt>
+                            <dd>0.6 vpred</dd>
+                            <dt>Base Model</dt>
+                            <dd>Illustrious</dd>
+                            <dt>Published</dt>
+                            <dd>2024-11-15</dd>
+                            <dt>Availability</dt>
+                            <dd>Public</dd>
+                            <dt>CivitAI Tags</dt>
+                            <dd>
+                                <div class="civitai-tags-container">
+                                    <span class="civitai-tag">anime</span><span class="civitai-tag">base model</span>
+                                </div>
+                            </dd>
+                            <dt>Download Link</dt>
+                            <dd><a href=https://civitai.com/api/download/models/1064242 target="_blank">https://civitai.com/api/download/models/1064242</a></dd>
+                        </dl>
+                        <div style="align-self:center; min-width:320px;">
+                            <div>
+                                <p style="line-height: 2; font-weight: bold;"><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Use the model without crediting the creator<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Sell images they generate<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Run on services that generate images for money<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Run on Civitai<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="lime"><path d="M5 12l5 5l10 -10"></path></svg></span> Share merges using this model<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="red"><path d="M18 6l-12 12"></path><path d="M6 6l12 12"></path></svg></span> Sell this model or merges using this model<br/><span style="display:inline-block;vertical-align:middle;"><svg width="15" height="15" viewBox="0 1.5 24 24" stroke-width="4" stroke-linecap="round" stroke="red"><path d="M18 6l-12 12"></path><path d="M6 6l12 12"></path></svg></span> Have different permissions when sharing merges</p>
+                            </div>
+                        </div>
+                    </div>
+                    <input type="checkbox" id="preview-civitai-description" class="description-toggle-checkbox">
+                    <div class="model-description">
+                        <h2>Description</h2>
+                        <h2 id="in-depth-retraining-of-illustrious-to-achieve-best-prompt-adherence-knowledge-and-state-of-the-art-performance.-66mudratg">In depth retraining of Illustrious to achieve best prompt adherence, knowledge and state of the art performance.</h2><p><span style="color:rgb(134, 142, 150)">Big dreams come true</span></p><p><span style="color:rgb(134, 142, 150)">The version number is just an index of current final release, not a fraction of the planned training.</span></p><p><a target="_blank" rel="ugc" href="https://huggingface.co/Minthy/RouWei-0.8"><span style="color:rgb(134, 142, 150)">HF repo</span></a></p><p></p><h3 id="large-scale-finetune-using-gpu-cluster-with-a-dataset-of-~13m-pictures-(~4m-with-natural-text-captions)-pqrwpp4f0">Large scale finetune using gpu cluster with a dataset of ~13M pictures (~4M with natural text captions)</h3><ul><li><p>Fresh and wast knowledge about characters, concepts, styles, cultural and related things</p></li><li><p>The best prompt adherence among SDXL anime models at the moment of release</p></li><li><p>Solved main problems with tags bleeding and biases, common for Illustrious, NoobAi and other checkpoints</p></li><li><p>Excellent aesthetics and knowledge across a wide range of styles (<a target="_blank" rel="ugc" href="https://files.catbox.moe/anleso.txt">over 50,000 artists</a> (<a rel="ugc" href="https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg">examples</a>), including hundreds of unique cherry-picked datasets from private galleries, including those received from the artists themselves)</p></li><li><p>High flexibility and variety without stability tradeoff</p></li><li><p>No more annoying watermarks for popular styles thanks to clean dataset</p></li><li><p>Vibrant colors and smooth gradients without trace of burning, full range even with epsilon</p></li><li><p>Pure training from Illustrious v0.1 without involving third-party checkpoints, Loras, tweakers, etc.</p></li></ul><p></p><h3 id="there-are-also-some-issues-and-changes-compared-to-the-previous-version-please-rtfm.-aaa28dbe6">There are also some issues and changes compared to the previous version, please RTFM.</h3><p></p><p>Dataset cut-off - end of April 2025.</p><p></p><h1 id="features-and-prompting:-j89sanl8e">Features and prompting:</h1><p></p><h2 id="important-change:-f6w18r0i0">Important change:</h2><p>When you are prompting artist styles, especially mixing several, <strong><u>their tags MUST BE in a separate CLIP chunk</u></strong>. Just add BREAK after it (for A1111 and derivatives), use conditioning concat node (for Comfy) <strong><u>or at least put them in the very end</u></strong>. Otherwise, significant degradation of results is likely.</p><p></p><h2 id="basic:-aroav4616">Basic:</h2><p></p><p>The checkpoint works both with short-simple and long-complex prompts. However, if there are contradictory or weird things - unlike with others they won't be ignored affecting the output. No guide-rails, no safeguards, no lobotomy.</p><p>Just prompt what you want to see and don't prompt what shouldn't be on the picture. If you want to have a view from above - don't put ceiling into positive, if you want to have crop view with head out of frame - don't make detailed description of character facial features, and so on. Pretty simple but sometimes people are missing it.</p><p></p><p>Version 0.8 comes with advanced understanding of natural text prompts. <u>It doesn't mean that you are obligated to use it</u>, tags only - completely fine, especially because understanding of tags combinations is also improved.</p><p>Do not expect it to perform like Flux or other models based on T5 or LLM text encoders. The whole size ot SDXL checkpoint is less then only that text encoder, in addition illustrious-v0.1 which is used as the base completely forgot a lot of general things from vanilla sdxl-base.</p><p>However, even in current state it works much better, allows to do new things usually impossible without external guidance, as well making manual editing, inpainting, etc more convenient.</p><p></p><p>To achieve best performance you should keep track of CLIP chunks. In SDXL the prompt is separated into a chunks of 75 (77 including BOS and EOS) tokens, that are processing by CLIP separately, and only then are concatinating and comes as conditions to unet.</p><p>If you want to specify some features for character/object and separate them from other prompt parts - make sure they are in the same chunk and optionally separate it with <code style="text-wrap: wrap">BREAK</code>. It will not solve problem of traits mixing completely, but can reduce it improving overall understanding, since text encoders on RouWei are able to process the whole sequence, not individual concepts better then others.</p><p>Dataset contains only booru-style tags and natural text expressions. Despite having a share of furries, real life photos, western media, etc. all captions have been converted to classic booru style to avoid a number of problems from mixing of different systems. So e621 tags won't be understanded properly.</p><p></p><h3 id="sampling-parameters:-7andwk13p">Sampling parameters:</h3><ul><li><p>~1 megapixel for txt2img, any AR with resolution multiple of 32 (1024x1024, 1056x, 1152x, 1216x832,...). Euler_a, 20..28steps.</p></li><li><p>CFG: for epsilon version 4..9 (7 is best), for vpred version, 3..5</p></li><li><p>Sigmas multiply may improve results a bit, CFG++ samplers work fine. LCM/PCM/DMD/... and exotic samplers untested.</p></li><li><p><strong>Some schedulers doesn't work well.</strong></p></li><li><p>Highresfix - x1.5 latent + denoise 0.6 or any gan + denoise 0.3..0.55.</p></li><li><p>For vpred version lower CFG 3..5 is needed!</p></li></ul><p></p><h3 id="for-vpred-version-lower-cfg-3..5-is-needed!-uwz9s21js"><strong>For vpred version lower CFG 3..5 is needed!</strong></h3><p></p><h2 id="quality-classification:-kw87tas72">Quality classification:</h2><p>Only 4 quality tags:</p><pre><code style="text-wrap: wrap">masterpiece, best quality</code></pre><p>for positive and</p><pre><code style="text-wrap: wrap">low quality, worst quality</code></pre><p>for negative.</p><p>Nothing else. Actually you can even omit positive and reduce negative to <code style="text-wrap: wrap">low quality</code> only, since they can affect basic style and composition.</p><p>Meta tags like <code style="text-wrap: wrap">lowres</code> have been removed and don't work, better not to use them. Low resolution images have been either removed or upscaled and cleaned with DAT depending on their importance.</p><p></p><h2 id="negative-prompt:-79eot1hhx">Negative prompt:</h2><pre><code style="text-wrap: wrap">worst quality, low quality, watermark</code></pre><p>That's all, no need of "rusty trombone", "farting on prey" and others. Do not put tags like greyscale, monochrome in negative unless you understand what are you doing. Extra tags for brightness/colors/contrast section below can be used</p><p></p><h2 id="artist-styles:-6yumq5hap">Artist styles:</h2><p></p><p><a target="_blank" rel="ugc" href="https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg">Grids with examples</a>, <a target="_blank" rel="ugc" href="https://files.catbox.moe/anleso.txt">list/wildcard</a> (also can be found in "training data").</p><p></p><h3 id="used-with-&quot;by-&quot;-it's-mandatory.-it-will-not-work-properly-without-it.-innc8cnap"><u>Used with "by " it's mandatory</u>. It will not work properly without it.</h3><p>"by " is a meta-token for styles to avoid mixing/misinterpret with tags/characters of similar or close name. This allows to have a better results for styles and at the same time avoid random style fluctuation that you may observe in other checkpoints.</p><p></p><p>Multiple give very interesting results, can be controlled with prompt weights and spells.</p><p></p><h3 id="you-must-add-break-after-artistsstyle-tags-(for-a1111)-or-concat-conditioning-(for-comfy)-or-put-them-in-the-very-end-of-your-prompt.-36i67w9c0">YOU MUST ADD <code style="text-wrap: wrap">BREAK</code> after artists/style tags (for A1111) or concat conditioning (for Comfy) or put them in the very end of your prompt.</h3><p>For example:</p><pre><code style="text-wrap: wrap">by kantoku, by wlop, best quality, masterpiece BREAK 1girl, ...</code></pre><p></p><h2 id="general-styles:-5ig9sgqza">General styles:</h2><pre><code style="text-wrap: wrap">2.5d, anime screencap, bold line, sketch, cgi, digital painting, flat colors, smooth shading, minimalistic, ink style, oil style, pastel style</code></pre><h3 id="booru-tags-styles:-uq05a7u5y">Booru tags styles:</h3><pre><code style="text-wrap: wrap">1950s (style), 1960s (style), 1970s (style), 1980s (style), 1990s (style), 2000s (style), animification, art nouveau, pinup (style), toon (style), western comics (style), nihonga, shikishi, minimalism, fine art parody</code></pre><p>and everything from <a target="_blank" rel="ugc" href="https://danbooru.donmai.us/wiki_pages/traditional_media">this group</a>.</p><p>Can be used in combinations (with artists too), with weights, both in positive and negative prompts.</p><p></p><h2 id="characters:-8htyrng9m">Characters:</h2><p>Use full name booru tag and proper formatting, like <code style="text-wrap: wrap">karin_(blue_archive)</code> -&gt; <code style="text-wrap: wrap">karin \(blue archive\)</code>, use skin tags for better reproducing, like <code style="text-wrap: wrap">karin \(bunny\) \(blue archive\)</code>. Autocomplete extension might be very useful.</p><p>Most characters are recognized just by their booru tag, but it will be more accurate if you describe their basic traits. Here you can easily redress your waifu/husbendo just by the prompt without suffering from the typical leaks of basic features.</p><p></p><h2 id="natural-text:-cq8whprlj">Natural text:</h2><p>Use it in combination with booru tags, works great. Use only natural text after typing styles and quality tags. Use just booru tags and forget about it, it's all up to you. To get best performance keep track if CLIP 75 tokens chunks.</p><p></p><p>About 4M of images in dataset had hybrid natural-text captions, made by Claude, GPT, Gemini, <a target="_blank" rel="ugc" href="https://huggingface.co/Minthy/ToriiGate-v0.4-7B">ToriiGate</a>, then refactored, cleaned and combined with tags in different variations for augmentation.</p><p></p><p>Unlike typical captions, these contains character names which is very useful. Better to keep it clean, short and convenient description works best. Better <strong><u>not</u></strong> use long and sloppy BS like</p><pre><code style="text-wrap: wrap">A mysteriously enchanting feminine entity of indeterminate yet youthful essence, whose celestial visage radiates with the ethereal luminescence of a thousand dying stars, blessed with locks cascading like the golden rivers of ancient mythology, perhaps styled in a manner reminiscent of contemporary fashion trends though not necessarily adhering to any specific aesthetic paradigm. Her eyes, pools of unfathomable depth and hue, sparkle with the wisdom of millennia yet maintain an innocent quality that defies temporal constraints...</code></pre><p>For captioning you can use <a target="_blank" rel="ugc" href="https://huggingface.co/Minthy/ToriiGate-v0.4-7B">ToriiGate</a> in short mode.</p><p></p><p>And don't expect it to be as good as flux and others, it tries very hard and after several rolls usually you can get what you want, but it is not that stable and detailed.</p><p></p><h2 id="lots-of-tailears-related-concepts:-qsxmk9gzm">Lots of Tail/Ears-related concepts:</h2><p>Oh yeah</p><pre><code style="text-wrap: wrap">tail censor, holding own tail, hugging own tail, holding another's tail, tail grab, tail raised, tail down, ears down, hand on own ear, tail around own leg, tail around penis, tailjob, tail through clothes, tail under clothes, lifted by tail, tail biting, tail penetration (including a specific indication of vaginal/anal), tail masturbation, holding with tail, panties on tail, bra on tail, tail focus, presenting own tail...</code></pre><p>(booru meaning, not e621) and many others with natural text. The majority works perfectly, some requires a lot of rolling.</p><p></p><h1 id="brightnesscolorscontrast:-lk2g7nwtj">Brightness/colors/contrast:</h1><p>You can use extra meta tags to control it:</p><pre><code style="text-wrap: wrap">low brightness, high brightness, low saturation, high saturation, low gamma, high gamma, sharp colors, soft colors, hdr, sdr</code></pre><p><a target="_blank" rel="ugc" href="https://files.catbox.moe/r8wrb6.jpg">Example</a></p><p>They work both in epsilon and vpred version and works really good.</p><p>Epsilon version relies on them too much. Without low brightness or low gamma or limited range (in negative) it might be difficult to achieve true 0,0,0 black, the same often true for white.</p><p>Both epsilon and vpred versions have like true zsnr, full range of colors and brightness without common flaws observed. But they behaves differently, just try it.</p><p></p><h1 id="vpred-version-9pr2w5h4d">Vpred version</h1><p></p><p>Main thing you need to know - lower your CFG from 7 down to 5 (or less). Otherwise, the use is similar with advantages.</p><p>It seems that starting from v0.7 vpred works flawlessly now. It shouldn't suffer from ignorance of tags close to the 75tokens chunk borders like nai. It is more difficult to get burned images - even on cfg7 usually it just over-saturated but with smooth gradients, which can be useful for some styles. Yes it can make anything from (0,0,0) to (255,255,255). You will find brightness meta tags described above quite useful for easier/lazy prompting, natural text expressions also work. To get the most dark image - put <code style="text-wrap: wrap">high brightness</code> into <u>negative</u> and/or use <code style="text-wrap: wrap">low brightness, low gamma</code> tags. If you don't like very bright skin on dark background and want to reduce contrast (or on the contrary, enhance the effect) - use hdr/sdr in negative/positive.</p><p>It was reported that in rare cases on some prompts there is a drop in contrast. Looks like other vpred models have same behaviour with such prompts, adding a "separator" closer to the border of the 75-token chunk fixes this. However, with 0.7 I haven't encountered this myself.</p><p></p><p><span style="color:rgb(134, 142, 150)">To launch vpred version you will need dev build of A1111, Comfy (with special loader node), Forge or Reforge. Just use same parameters (Euler a, cfg 3..5, 20..28 steps) like epsilon. No need to use Cfg rescale, but you can try it, cfg++ works great.</span></p><p></p><h1 id="base-model:-lwa0dfdbf">Base model:</h1><p>The model here has a small unet polishint after main training to improve small details, bump up resolution and others. Hovewer, you may be also interested into a RouWei-Base, which sometimes can perform better at complex prompts despite having minor mistakes in small details. It also comes in FP32, for example if you want to use fp32 text encoder nodes in Comfy, merge it or finetune.</p><p>It can be found in <a target="_blank" rel="ugc" href="https://huggingface.co/Minthy/RouWei-0.8">Huggingface repo</a></p><p></p><h1 id="known-issues:-jxgt6yemu">Known issues:</h1><p>Off course there are:</p><ul><li><p>Artists and style tags must be seperated into a different chunk from main prompt or come very last</p></li><li><p>There may be some positional or combinational bias in rare cases, but it's not yet clear.</p></li><li><p>There are some complaints about few of the general styles.</p></li><li><p>Epsilon version relies too much on brightness meta tags, sometimes you will need to use them to get desired brightness shift</p></li><li><p>Some newly added styles/characters might be not as good and disctinct as they deserve to</p></li><li><p>To be discovered</p><p></p></li><li><p></p></li></ul><h3 id="requests-for-artistscharacters-in-future-models-are-open.-if-you-find-artistcharacterconcept-that-perform-weak-inaccurate-or-has-strong-watermark-please-report-will-add-them-explicitly.-follow-for-a-new-versions.-kgc39ed4e">Requests for artists/characters in future models are open. If you find artist/character/concept that perform weak, inaccurate or has strong watermark - please report, will add them explicitly. Follow for a new versions.</h3><p></p><h2 id="join-the-discord-server-9lpaojolv"><a target="_blank" rel="ugc" href="https://discord.gg/ZXHENAhqE9">JOIN THE DISCORD SERVER</a></h2><p></p><h2 id="license:-0hpy342uh">License:</h2><p>Same as illustrious. Fell free to use in your merges, finetunes, ets. but please leave a link or mention, it is mandatory</p><p></p><h2 id="how-it's-made-zc68i8qve">How it's made</h2><p>I'll consider to make a report or something like it later. For sure.</p><p>In short, 98% of work is related to dataset preparations. Instead of blindly relying on loss-weighting based on tag frequency from nai paper, a custom guided loss-weighting implementation along with asynchronous collator for balancing have been used. Ztsnr (or close to it) with Epsilon prediction was achieved using noise scheduler augmentation.</p><p></p><p>Spent compute - over 8k hours of H100 (apart from research and fail attempts)</p><p></p><h1 id="thanks:-4tsy2pj9q">Thanks:</h1><p>First of all I'd like to acknowledge everyone who supports open source, develops in improves code. Thanks to the authors of illustrious for releasing model, thank to NoobAI team for being pioneers in open finetuning of such a scale, sharing experience, raising and solving issues that previously went unnoticed.</p><h3 id="personal:-9a91vqx2i">Personal:</h3><p>Artists wish to remain anonymous for sharing private works; Few anonymous persons - donations, code, captions, etc., Soviet Cat - GPU sponsoring; Sv1. - llm access, captioning, code; K. - training code; Bakariso - datasets, testing, advices, insides; NeuroSenko - donations, testing, code; LOL2024 - a lot of unique datasets; T.,[] - datasets, testing, advises; rred, dga, Fi., ello - donations; TekeshiX - datasets. And other fellow brothers that helped. Love you so much ❤️.</p><p>And off course everyone who made feedback and requests, it's really valuable.</p><p><span style="color:rgb(134, 142, 150)">If I forgot to mention anyone, please notify.</span></p><p></p><h2 id="donations-jdo27wd0z">Donations</h2><p>If you want to support - share my models, leave feedback, make a cute picture with kemonomimi-girl. And of course, support original artists.</p><p>AI is my hobby, I'm spending money on it and not begging for donations. However, it has turned into a large-scale and expensive undertaking. Consider to support to accelerate new training and researches.</p><p><span style="color:rgb(134, 142, 150)">(Just keep in mind that I can waste it on alcohol or cosplay girls)</span></p><p>BTC: bc1qwv83ggq8rvv07uk6dv4njs0j3yygj3aax4wg6c</p><p>ETH/USDT(e): 0x04C8a749F49aE8a56CB84cF0C99CD9E92eDB17db</p><p>XMR: 47F7JAyKP8tMBtzwxpoZsUVB8wzg2VrbtDKBice9FAS1FikbHEXXPof4PAb42CQ5ch8p8Hs4RvJuzPHDtaVSdQzD6ZbA5TZ</p><p>if you can offer gpu-time (a100+) - PM.</p><p></p><p></p><p></p><p></p><p></p>
+                    </div>
+                    <label for="preview-civitai-description" class="description-toggle-label"></label>
+                </div>
+                <div align=center><div class="sampleimgs"><input type="radio" name="zoomRadio" id="resetZoom" class="zoom-radio" checked>
+                    <div class="model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_0" class="zoom-radio">
+                    <label for="zoomRadiopreview_0" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/f96c7ce9-0eaf-4461-bee3-176a008ee8d2/width=1536/40258578.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/f96c7ce9-0eaf-4461-bee3-176a008ee8d2/width=1536/40258578.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>masterpiece, best quality, good quality,
+1girl, konpaku youmu, half-closed eyes, solo, black hairband, (konpaku youmu (ghost):1.2), hairband, green vest, puffy sleeves, upper body, bow, black bowtie, black bow, short hair, vest, bowtie, ribbon, leaf, shirt, wind, short sleeves, black background, looking down, white shirt, hitodama, black ribbon, hair ribbon, puffy short sleeves, grey hair, simple background, floating hair, closed mouth, expressionless, wing collar, hitodama print, buttons, green eyes, collared shirt, falling leaves, aqua eyes, glowing eyes, pale skin, (skeleton, split screen, straight-on, android, crank, gauge, gears, glowing, light bulb, mechabare, sparkle, x-ray:1.2), 
+&lt;lora:youmuanon_noob-vpred0.1_flawless_ds_v2:1&gt;</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, bad quality, lowres, jpeg artifacts, bad anatomy, bad hands, multiple views, signature, watermark, censored, covered navel, limited range, high brightness, high contrast,</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>4147779862</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1024x1024</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred_rc1</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>5</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_0">
+                                    <label class="tab-label" for="chckpreview_0">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl-vae-fp16-fix.safetensors</dd></div><div class="civitai-meta"><dt>Ensd</dt><dd>31337</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;62c7c729ad&#x27;, &#x27;model&#x27;: &#x27;e63d568c63&#x27;, &#x27;lora:youmuanon_noob-vpred0.1_flawless_ds_v2&#x27;: &#x27;7b7e860a9bae&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.3.6dev2-v1.10.1RC-latest-1453-g0fd79513</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;7b7e860a9bae&#x27;, &#x27;name&#x27;: &#x27;youmuanon_noob-vpred0.1_flawless_ds_v2&#x27;, &#x27;type&#x27;: &#x27;lora&#x27;, &#x27;weight&#x27;: 1}, {&#x27;hash&#x27;: &#x27;e63d568c63&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred_rc1&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>e63d568c63</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Align Your Steps</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>2x-AnimeSharpV3</dd></div><div class="civitai-meta"><dt>Adetailer model</dt><dd>face_yolov8n.pt</dd></div><div class="civitai-meta"><dt>Adetailer version</dt><dd>24.9.0</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Sigmas_merge_mode</dt><dd>Multiply</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.5</dd></div><div class="civitai-meta"><dt>Sigmas_mult_factor</dt><dd>2</dd></div><div class="civitai-meta"><dt>Adetailer mask blur</dt><dd>8</dd></div><div class="civitai-meta"><dt>Adetailer confidence</dt><dd>0.3</dd></div><div class="civitai-meta"><dt>Sigmas_merge_enabled</dt><dd>True</dd></div><div class="civitai-meta"><dt>Adetailer dilate erode</dt><dd>4</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Adetailer inpaint padding</dt><dd>32</dd></div><div class="civitai-meta"><dt>Adetailer negative prompt</dt><dd>{}</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div><div class="civitai-meta"><dt>Adetailer denoising strength</dt><dd>0.4</dd></div><div class="civitai-meta"><dt>Old prompt editing timelines</dt><dd>True</dd></div><div class="civitai-meta"><dt>Adetailer inpaint only masked</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_1" class="zoom-radio">
+                    <label for="zoomRadiopreview_1" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/76198546-76cd-4bf4-9fc8-f935da633bbf/width=1328/40258585.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/76198546-76cd-4bf4-9fc8-f935da633bbf/width=1328/40258585.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>(by deel \(rkeg\):0.8), best quality, masterpiece,
+1girl, folinic \(arknights\), cat girl, brown hair, multicolored hair, cat ears, surprised, blush, hair flower, holding papers, BREAK
+indoors, scientific lab, glass floor, intricate interior, computers, research facility</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, watermark</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>2290774044</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>832x1216</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_1">
+                                    <label class="tab-label" for="chckpreview_1">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;84c8888446&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;84c8888446&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>84c8888446</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.6</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_2" class="zoom-radio">
+                    <label for="zoomRadiopreview_2" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/5033de43-81ec-418e-9429-f2a7425353c4/width=1728/40258592.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/5033de43-81ec-418e-9429-f2a7425353c4/width=1728/40258592.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by ratatatat74, by tarakanovich,cube85, masterpiece, best quality,
+1girl, fake animal ears, cat ears, long black hair, large breasts, erect nipples, completely nude, butt plug, black thighhighs, garter straps, squatting, reverse cowgirl position, girl on top, from behind, ass, mature female, pussy juice, ass grab, large penis, vaginal, pov hands, tally, (slap mark:0.7),
+indoors, bedroom, dark night, dimly lit room, dark blue theme, low brightness, low gamma, sdr</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, limited range</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>1178610355</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1152x896</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_2">
+                                    <label class="tab-label" for="chckpreview_2">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_3" class="zoom-radio">
+                    <label for="zoomRadiopreview_3" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/9af10cd7-4513-4147-a4f2-033f46afd3e0/width=1824/40258577.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/9af10cd7-4513-4147-a4f2-033f46afd3e0/width=1824/40258577.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>(by ratatatat74:0.7), by tarakanovich, cube85, masterpiece, best quality,
+1girl, vulpisfoglia \(arknights\), fox girl, fox ears, large ears, ears down, blonde hair, large breasts, yellow eyes, fox tail, nightgown, see-through, bottomless, from behind, tail censor, (scar on animal ear:0.8), white panties, naughty face,
+all fours, fox shadow puppet, glowing eyes,
+black background, low brightness, low gamma</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, extra ears</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>1556719420</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1216x832</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>20</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_3">
+                                    <label class="tab-label" for="chckpreview_3">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_4" class="zoom-radio">
+                    <label for="zoomRadiopreview_4" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/0d48a8dd-f3d3-42f6-b79d-de7de7b52139/width=1200/40258590.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/0d48a8dd-f3d3-42f6-b79d-de7de7b52139/width=1200/40258590.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>masterpiece, best quality, by merrytail, by sub-res, by kie \(wylee2212\),
+1girl, fox devil \(chainsaw man\), fox girl, fox ears, fox tail, whisker markings, long white hair, red eyes, (extra eyes:1.3), fangs, evil smile, white kimono with floral print, collarbone, hair bun, fox shadow puppet, cowboy shot, posing,
+black background, low brightness</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, limited range, monster, 1other, animal, fox</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>1376750140</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>800x1344</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_4">
+                                    <label class="tab-label" for="chckpreview_4">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_5" class="zoom-radio">
+                    <label for="zoomRadiopreview_5" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/209ff688-e01d-41d4-b842-a3ffadd3741f/width=1824/40258575.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/209ff688-e01d-41d4-b842-a3ffadd3741f/width=1824/40258575.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>masterpiece, best quality, by done \(donezumi\),
+1girl, kon \(tokyo ravens\), tokyo ravens, fox girl, fox ears, fox tail, white hair, thick eyebrows, pale skin, blue eyes, white aqua japanese clothes, detached sleeves, ribbon trim, fox shadow puppet, pom pom \(clothes\), black shorts, white thighhighs, hips, happy, BREAK floating in air, legs apart, posing, full body, wide shot,
+white background, high brightness</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>2168859150</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1216x832</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>20</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_5">
+                                    <label class="tab-label" for="chckpreview_5">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;84c8888446&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;84c8888446&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>84c8888446</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_6" class="zoom-radio">
+                    <label for="zoomRadiopreview_6" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/90435b3a-ef7d-4be2-b700-00601088c553/width=1248/40258594.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/90435b3a-ef7d-4be2-b700-00601088c553/width=1248/40258594.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by tandohark, (by personal ami:0.6), best quality, masterpiece,
+2girls, asymmetrical docking BREAK
+izuna \(blue archive\), fox girl, fox ears, fox mask, fox tail, halo, topless, denim shorts, striped bikini bra, medium breasts, sideboob, visors cap BREAK
+wakamo \(blue archive\), fox girl, fox ears, fox tail, long black hair, red halo, hair ornament, large breasts, fox mask, white bikini swimsuit, collarbone, :3 BREAK
+outdoors, beach, seashore, palms, clear blue sky</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, watermark</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>1114850771</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>832x1216</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>5</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_6">
+                                    <label class="tab-label" for="chckpreview_6">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;84c8888446&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;84c8888446&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>84c8888446</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_7" class="zoom-radio">
+                    <label for="zoomRadiopreview_7" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/23c627f5-451b-47de-a1d3-7dce53adb8d9/width=1248/40258589.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/23c627f5-451b-47de-a1d3-7dce53adb8d9/width=1248/40258589.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by akai sashimi, by nat the lich, by foxykuro, by nekoya \(liu\), masterpiece,  best quality,
+1girl, kal&#x27;tsit \(arknights\), cat girl, lynx ears, silver hair, green eyes/, naked towel, covered nipples, off shoulder, oripathy lesion \(arknights\), collarbone, shaded face, (evil smile:1.1), smug, open mouth, angry, fang, tilted head, wet,
+holding handcuffs,
+cowboy shot, motion lines,
+indoors, office, dimmed, dark background</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, watermark, dutch angle</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>2542436241</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>832x1216</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>28</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_7">
+                                    <label class="tab-label" for="chckpreview_7">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Ensd</dt><dd>31337</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Discrete_zsnr</dt><dd>True</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Discrete_sampling</dt><dd>v_prediction</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Advanced_sampling_mode</dt><dd>Discrete</dd></div><div class="civitai-meta"><dt>Advanced_sampling_enabled</dt><dd>True</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_8" class="zoom-radio">
+                    <label for="zoomRadiopreview_8" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/fd31b937-db26-404f-ba61-c86eecedb37c/width=1328/40258587.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/fd31b937-db26-404f-ba61-c86eecedb37c/width=1328/40258587.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by akai sashimi, by nat the lich, by foxykuro, by nekoya \(liu\), masterpiece, best quality,
+1girl, kal&#x27;tsit \(arknights\), cat girl, lynx ears, silver hair, green eyes, short tail, completely nude, oripathy lesion \(arknights\), scars on spine, reverse cowgirl position, motion lines, from below, open mouth, smug, heavenly ass, v,
+1boy, vaginal, sex, pov, deep penetration, pussy juice trail,
+indoors, bedroom, dimmed, dark background</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, watermark, abstract background</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>3956367149</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>832x1216</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>20</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_8">
+                                    <label class="tab-label" for="chckpreview_8">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;21e24eab73&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;21e24eab73&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>21e24eab73</dd></div><div class="civitai-meta"><dt>Hires steps</dt><dd>18</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.6</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div></div></div></div></dl></div></div>
+                    <div class="civnsfw model-block" style="display:flex;align-items:flex-start;">
+                    <div class="civitai-image-container">
+                    <input type="radio" name="zoomRadio" id="zoomRadiopreview_9" class="zoom-radio">
+                    <label for="zoomRadiopreview_9" class="zoom-img-container">
+                    <img data-sampleimg="true" src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/11c49fc0-f6fa-40b2-bb86-30e3a2c9cd0a/width=2064/40258599.jpeg">
+                        </label>
+                        <label for="resetZoom" class="zoom-overlay"></label>
+                    
+                            <div class="civitai_txt2img" style="margin-top:30px;margin-bottom:30px;">
+                            <label onclick='sendImgUrl("https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/11c49fc0-f6fa-40b2-bb86-30e3a2c9cd0a/width=2064/40258599.jpeg")' class="civitai-txt2img-btn" style="max-width:fit-content;cursor:pointer;">Send to txt2img</label>
+                            </div></div>
+                        <div style="margin:1em 0em 1em 1em;text-align:left;line-height:1.5em;" id="image_info"><dl style="gap:10px; display:grid;"><div class="civitai-meta-btn" onclick="metaToTxt2Img('Prompt', this)"><dt>Prompt</dt><dd>by kate fox, masterpiece, best quality,
+1girl, yuel \(granblue fantasy\), fox girl, yellow eyes, fox ears, large fox tail, bikini bra, fur trim, elbow gloves, :3, wide shot, BREAK
+outdoors, torii, asian nature, epic landscape of mountains, above clouds</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Negative prompt', this)"><dt>Negative prompt</dt><dd>worst quality, low quality, bad hands, watermark
+worst quality, low quality</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Seed', this)"><dt>Seed</dt><dd>2135703813</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Size', this)"><dt>Size</dt><dd>1376x720</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Model', this)"><dt>Model</dt><dd>rouwei_0.6_vpred</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Sampler', this)"><dt>Sampler</dt><dd>Euler a</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('Steps', this)"><dt>Steps</dt><dd>20</dd></div><div class="civitai-meta-btn" onclick="metaToTxt2Img('CFG scale', this)"><dt>CFG scale</dt><dd>7</dd></div>
+                            <div class="tabs">
+                                <div class="tab">
+                                    <input type="checkbox" class="accordionCheckbox" id="chckpreview_9">
+                                    <label class="tab-label" for="chckpreview_9">More details...</label>
+                                    <div class="tab-content" style="gap:10px;display:grid;margin-left:1px;">
+                            <div class="civitai-meta"><dt>Vae</dt><dd>sdxl_vae.safetensors</dd></div><div class="civitai-meta"><dt>Hashes</dt><dd>{&#x27;vae&#x27;: &#x27;235745af8d&#x27;, &#x27;model&#x27;: &#x27;84c8888446&#x27;}</dd></div><div class="civitai-meta"><dt>Version</dt><dd>f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32</dd></div><div class="civitai-meta"><dt>Emphasis</dt><dd>No norm</dd></div><div class="civitai-meta"><dt>Resources</dt><dd>[{&#x27;hash&#x27;: &#x27;84c8888446&#x27;, &#x27;name&#x27;: &#x27;rouwei_0.6_vpred&#x27;, &#x27;type&#x27;: &#x27;model&#x27;}]</dd></div><div class="civitai-meta"><dt>Model hash</dt><dd>84c8888446</dd></div><div class="civitai-meta"><dt>Hires upscale</dt><dd>1.5</dd></div><div class="civitai-meta"><dt>Schedule type</dt><dd>Automatic</dd></div><div class="civitai-meta"><dt>Hires upscaler</dt><dd>Latent</dd></div><div class="civitai-meta"><dt>Denoising strength</dt><dd>0.6</dd></div><div class="civitai-meta"><dt>Rescale_cfg_enabled</dt><dd>True</dd></div><div class="civitai-meta"><dt>Rescale_cfg_multiplier</dt><dd>0.7</dd></div></div></div></div></dl></div></div></div></div>
+                

Stable-diffusion/rouwei_v080Vpred_trainingData.json

@@ -0,0 +1,8 @@
+{
+    "sha256": "99042258AA3E20F4309C669E80FAC53326DC246036EE449CC47EAB0F6DBFE11B",
+    "modelId": 950531,
+    "modelVersionId": 1832460,
+    "activation text": "",
+    "description": "In depth retraining of Illustrious to achieve best prompt adherence, knowledge and state of the art performance.Big dreams come trueThe version number is just an index of current final release, not a fraction of the planned training.HF repo https://huggingface.co/Minthy/RouWei-0.8Large scale finetune using gpu cluster with a dataset of ~13M pictures (~4M with natural text captions)Fresh and wast knowledge about characters, concepts, styles, cultural and related thingsThe best prompt adherence among SDXL anime models at the moment of releaseSolved main problems with tags bleeding and biases, common for Illustrious, NoobAi and other checkpointsExcellent aesthetics and knowledge across a wide range of styles (over 50,000 artists https://files.catbox.moe/anleso.txt (examples https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg), including hundreds of unique cherry-picked datasets from private galleries, including those received from the artists themselves)High flexibility and variety without stability tradeoffNo more annoying watermarks for popular styles thanks to clean datasetVibrant colors and smooth gradients without trace of burning, full range even with epsilonPure training from Illustrious v0.1 without involving third-party checkpoints, Loras, tweakers, etc.There are also some issues and changes compared to the previous version, please RTFM.Dataset cut-off - end of April 2025.Features and prompting:Important change:When you are prompting artist styles, especially mixing several, their tags MUST BE in a separate CLIP chunk. Just add BREAK after it (for A1111 and derivatives), use conditioning concat node (for Comfy) or at least put them in the very end. Otherwise, significant degradation of results is likely.Basic:The checkpoint works both with short-simple and long-complex prompts. However, if there are contradictory or weird things - unlike with others they won't be ignored affecting the output. No guide-rails, no safeguards, no lobotomy.Just prompt what you want to see and don't prompt what shouldn't be on the picture. If you want to have a view from above - don't put ceiling into positive, if you want to have crop view with head out of frame - don't make detailed description of character facial features, and so on. Pretty simple but sometimes people are missing it.Version 0.8 comes with advanced understanding of natural text prompts. It doesn't mean that you are obligated to use it, tags only - completely fine, especially because understanding of tags combinations is also improved.Do not expect it to perform like Flux or other models based on T5 or LLM text encoders. The whole size ot SDXL checkpoint is less then only that text encoder, in addition illustrious-v0.1 which is used as the base completely forgot a lot of general things from vanilla sdxl-base.However, even in current state it works much better, allows to do new things usually impossible without external guidance, as well making manual editing, inpainting, etc more convenient.To achieve best performance you should keep track of CLIP chunks. In SDXL the prompt is separated into a chunks of 75 (77 including BOS and EOS) tokens, that are processing by CLIP separately, and only then are concatinating and comes as conditions to unet.If you want to specify some features for character/object and separate them from other prompt parts - make sure they are in the same chunk and optionally separate it with BREAK. It will not solve problem of traits mixing completely, but can reduce it improving overall understanding, since text encoders on RouWei are able to process the whole sequence, not individual concepts better then others.Dataset contains only booru-style tags and natural text expressions. Despite having a share of furries, real life photos, western media, etc. all captions have been converted to classic booru style to avoid a number of problems from mixing of different systems. So e621 tags won't be understanded properly.Sampling parameters:~1 megapixel for txt2img, any AR with resolution multiple of 32 (1024x1024, 1056x, 1152x, 1216x832,...). Euler_a, 20..28steps.CFG: for epsilon version 4..9 (7 is best), for vpred version, 3..5Sigmas multiply may improve results a bit, CFG++ samplers work fine. LCM/PCM/DMD/... and exotic samplers untested.Some schedulers doesn't work well.Highresfix - x1.5 latent + denoise 0.6 or any gan + denoise 0.3..0.55.For vpred version lower CFG 3..5 is needed!For vpred version lower CFG 3..5 is needed!Quality classification:Only 4 quality tags:masterpiece, best qualityfor positive andlow quality, worst qualityfor negative.Nothing else. Actually you can even omit positive and reduce negative to low quality only, since they can affect basic style and composition.Meta tags like lowres have been removed and don't work, better not to use them. Low resolution images have been either removed or upscaled and cleaned with DAT depending on their importance.Negative prompt:worst quality, low quality, watermarkThat's all, no need of \"rusty trombone\", \"farting on prey\" and others. Do not put tags like greyscale, monochrome in negative unless you understand what are you doing. Extra tags for brightness/colors/contrast section below can be usedArtist styles:Grids with examples https://mega.nz/folder/oSoDwIJZ#lZfjw20CT4W4ONcu_O8fjg, list/wildcard https://files.catbox.moe/anleso.txt (also can be found in \"training data\").Used with \"by \" it's mandatory. It will not work properly without it.\"by \" is a meta-token for styles to avoid mixing/misinterpret with tags/characters of similar or close name. This allows to have a better results for styles and at the same time avoid random style fluctuation that you may observe in other checkpoints.Multiple give very interesting results, can be controlled with prompt weights and spells.YOU MUST ADD BREAK after artists/style tags (for A1111) or concat conditioning (for Comfy) or put them in the very end of your prompt.For example:by kantoku, by wlop, best quality, masterpiece BREAK 1girl, ...General styles:2.5d, anime screencap, bold line, sketch, cgi, digital painting, flat colors, smooth shading, minimalistic, ink style, oil style, pastel styleBooru tags styles:1950s (style), 1960s (style), 1970s (style), 1980s (style), 1990s (style), 2000s (style), animification, art nouveau, pinup (style), toon (style), western comics (style), nihonga, shikishi, minimalism, fine art parodyand everything from this group https://danbooru.donmai.us/wiki_pages/traditional_media.Can be used in combinations (with artists too), with weights, both in positive and negative prompts.Characters:Use full name booru tag and proper formatting, like karin_(blue_archive) -> karin \\(blue archive\\), use skin tags for better reproducing, like karin \\(bunny\\) \\(blue archive\\). Autocomplete extension might be very useful.Most characters are recognized just by their booru tag, but it will be more accurate if you describe their basic traits. Here you can easily redress your waifu/husbendo just by the prompt without suffering from the typical leaks of basic features.Natural text:Use it in combination with booru tags, works great. Use only natural text after typing styles and quality tags. Use just booru tags and forget about it, it's all up to you. To get best performance keep track if CLIP 75 tokens chunks.About 4M of images in dataset had hybrid natural-text captions, made by Claude, GPT, Gemini, ToriiGate https://huggingface.co/Minthy/ToriiGate-v0.4-7B, then refactored, cleaned and combined with tags in different variations for augmentation.Unlike typical captions, these contains character names which is very useful. Better to keep it clean, short and convenient description works best. Better not use long and sloppy BS likeA mysteriously enchanting feminine entity of indeterminate yet youthful essence, whose celestial visage radiates with the ethereal luminescence of a thousand dying stars, blessed with locks cascading like the golden rivers of ancient mythology, perhaps styled in a manner reminiscent of contemporary fashion trends though not necessarily adhering to any specific aesthetic paradigm. Her eyes, pools of unfathomable depth and hue, sparkle with the wisdom of millennia yet maintain an innocent quality that defies temporal constraints...For captioning you can use ToriiGate https://huggingface.co/Minthy/ToriiGate-v0.4-7B in short mode.And don't expect it to be as good as flux and others, it tries very hard and after several rolls usually you can get what you want, but it is not that stable and detailed.Lots of Tail/Ears-related concepts:Oh yeahtail censor, holding own tail, hugging own tail, holding another's tail, tail grab, tail raised, tail down, ears down, hand on own ear, tail around own leg, tail around penis, tailjob, tail through clothes, tail under clothes, lifted by tail, tail biting, tail penetration (including a specific indication of vaginal/anal), tail masturbation, holding with tail, panties on tail, bra on tail, tail focus, presenting own tail...(booru meaning, not e621) and many others with natural text. The majority works perfectly, some requires a lot of rolling.Brightness/colors/contrast:You can use extra meta tags to control it:low brightness, high brightness, low saturation, high saturation, low gamma, high gamma, sharp colors, soft colors, hdr, sdrExampleThey work both in epsilon and vpred version and works really good.Epsilon version relies on them too much. Without low brightness or low gamma or limited range (in negative) it might be difficult to achieve true 0,0,0 black, the same often true for white.Both epsilon and vpred versions have like true zsnr, full range of colors and brightness without common flaws observed. But they behaves differently, just try it.Vpred versionMain thing you need to know - lower your CFG from 7 down to 5 (or less). Otherwise, the use is similar with advantages.It seems that starting from v0.7 vpred works flawlessly now. It shouldn't suffer from ignorance of tags close to the 75tokens chunk borders like nai. It is more difficult to get burned images - even on cfg7 usually it just over-saturated but with smooth gradients, which can be useful for some styles. Yes it can make anything from (0,0,0) to (255,255,255). You will find brightness meta tags described above quite useful for easier/lazy prompting, natural text expressions also work. To get the most dark image - put high brightness into negative and/or use low brightness, low gamma tags. If you don't like very bright skin on dark background and want to reduce contrast (or on the contrary, enhance the effect) - use hdr/sdr in negative/positive.It was reported that in rare cases on some prompts there is a drop in contrast. Looks like other vpred models have same behaviour with such prompts, adding a \"separator\" closer to the border of the 75-token chunk fixes this. However, with 0.7 I haven't encountered this myself.To launch vpred version you will need dev build of A1111, Comfy (with special loader node), Forge or Reforge. Just use same parameters (Euler a, cfg 3..5, 20..28 steps) like epsilon. No need to use Cfg rescale, but you can try it, cfg++ works great.Base model:The model here has a small unet polishint after main training to improve small details, bump up resolution and others. Hovewer, you may be also interested into a RouWei-Base, which sometimes can perform better at complex prompts despite having minor mistakes in small details. It also comes in FP32, for example if you want to use fp32 text encoder nodes in Comfy, merge it or finetune.It can be found in Huggingface repo https://huggingface.co/Minthy/RouWei-0.8Known issues:Off course there are:Artists and style tags must be seperated into a different chunk from main prompt or come very lastThere may be some positional or combinational bias in rare cases, but it's not yet clear.There are some complaints about few of the general styles.Epsilon version relies too much on brightness meta tags, sometimes you will need to use them to get desired brightness shiftSome newly added styles/characters might be not as good and disctinct as they deserve toTo be discoveredRequests for artists/characters in future models are open. If you find artist/character/concept that perform weak, inaccurate or has strong watermark - please report, will add them explicitly. Follow for a new versions.JOIN THE DISCORD SERVER https://discord.gg/ZXHENAhqE9License:Same as illustrious. Fell free to use in your merges, finetunes, ets. but please leave a link or mention, it is mandatoryHow it's madeI'll consider to make a report or something like it later. For sure.In short, 98% of work is related to dataset preparations. Instead of blindly relying on loss-weighting based on tag frequency from nai paper, a custom guided loss-weighting implementation along with asynchronous collator for balancing have been used. Ztsnr (or close to it) with Epsilon prediction was achieved using noise scheduler augmentation.Spent compute - over 8k hours of H100 (apart from research and fail attempts)Thanks:First of all I'd like to acknowledge everyone who supports open source, develops in improves code. Thanks to the authors of illustrious for releasing model, thank to NoobAI team for being pioneers in open finetuning of such a scale, sharing experience, raising and solving issues that previously went unnoticed.Personal:Artists wish to remain anonymous for sharing private works; Few anonymous persons - donations, code, captions, etc., Soviet Cat - GPU sponsoring; Sv1. - llm access, captioning, code; K. - training code; Bakariso - datasets, testing, advices, insides; NeuroSenko - donations, testing, code; LOL2024 - a lot of unique datasets; T.,[] - datasets, testing, advises; rred, dga, Fi., ello - donations; TekeshiX - datasets. And other fellow brothers that helped. Love you so much \u2764\ufe0f.And off course everyone who made feedback and requests, it's really valuable.If I forgot to mention anyone, please notify.DonationsIf you want to support - share my models, leave feedback, make a cute picture with kemonomimi-girl. And of course, support original artists.AI is my hobby, I'm spending money on it and not begging for donations. However, it has turned into a large-scale and expensive undertaking. Consider to support to accelerate new training and researches.(Just keep in mind that I can waste it on alcohol or cosplay girls)BTC: bc1qwv83ggq8rvv07uk6dv4njs0j3yygj3aax4wg6cETH/USDT(e): 0x04C8a749F49aE8a56CB84cF0C99CD9E92eDB17dbXMR: 47F7JAyKP8tMBtzwxpoZsUVB8wzg2VrbtDKBice9FAS1FikbHEXXPof4PAb42CQ5ch8p8Hs4RvJuzPHDtaVSdQzD6ZbA5TZif you can offer gpu-time (a100+) - PM.",
+    "sd version": "Other"
+}

Stable-diffusion/rouwei_v080Vpred_trainingData.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:99042258aa3e20f4309c669e80fac53326dc246036ee449cc47eab0f6dbfe11b
+size 682418