diff --git "a/Stable-diffusion/rouwei_v07Epsilon.html" "b/Stable-diffusion/rouwei_v07Epsilon.html" --- "a/Stable-diffusion/rouwei_v07Epsilon.html" +++ "b/Stable-diffusion/rouwei_v07Epsilon.html" @@ -6,15 +6,15 @@

RouWei

-

Uploaded by Minthybasis

+

Uploaded by Minthybasis

Version
-
v0.7 epsilon
+
0.6 vpred
Base Model
Illustrious
Published
-
2025-01-26
+
2024-11-15
Availability
Public
CivitAI Tags
@@ -24,224 +24,227 @@
Download Link
-
https://civitai.com/api/download/models/1330983
+
https://civitai.com/api/download/models/1064242
-

Use the model without crediting the creator
Sell images they generate
Run on services that generate images for money
Run on Civitai
Share merges using this model
Sell this model or merges using this model
Have different permissions when sharing merges

+

Use the model without crediting the creator
Sell images they generate
Run on services that generate images for money
Run on Civitai
Share merges using this model
Sell this model or merges using this model
Have different permissions when sharing merges

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):

v0.7 vpred will be released soon

Key advantages:

An addition to mentioned, comparing with vanilla Illustrious and NoobAI:

Dataset 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:

For vpred version lower CFG 3..5 is needed!

Quality classification:

Only 4 quality tags:

masterpiece, best quality
low quality, worst quality

Nothing 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, watermark

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. 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, list (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 style

Booru 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 parody

and everything from this group.

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, 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.

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, sdr

Example

They 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 version

Vpred 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:

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.

JOIN THE DISCORD SERVER

License:

Same as illustrious. Fell free to use in your merges, finetunes, ets. just please leave a link.

How it's made

I'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 ❤️.

And off course everyone who made feedback and requests, it's really valuable.

If I forgot to mention anyone, please notify.

Donations

If 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: bc1qwv83ggq8rvv07uk6dv4njs0j3yygj3aax4wg6c

ETH/USDT(e): 0x04C8a749F49aE8a56CB84cF0C99CD9E92eDB17db

XMR: 47F7JAyKP8tMBtzwxpoZsUVB8wzg2VrbtDKBice9FAS1FikbHEXXPof4PAb42CQ5ch8p8Hs4RvJuzPHDtaVSdQzD6ZbA5TZ

if you can offer gpu-time (a100+) - PM.

+

In depth retraining of Illustrious to achieve best prompt adherence, knowledge and state of the art performance.

Big dreams come true

The version number is just an index of current final release, not a fraction of the planned training.

HF repo

Large scale finetune using gpu cluster with a dataset of ~13M pictures (~4M with natural text captions)

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:

For vpred version lower CFG 3..5 is needed!

Quality classification:

Only 4 quality tags:

masterpiece, best quality

for positive and

low quality, worst quality

for 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, watermark

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

Artist styles:

Grids with examples, list/wildcard (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 style

Booru 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 parody

and everything from this group.

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, 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 like

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...

For captioning you can use ToriiGate 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.

Oh yeah

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...

(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, sdr

Example

They 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 version

Main 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

Known issues:

Off course there are:

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.

JOIN THE DISCORD SERVER

License:

Same as illustrious. Fell free to use in your merges, finetunes, ets. but please leave a link or mention, it is mandatory

How it's made

I'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 ❤️.

And off course everyone who made feedback and requests, it's really valuable.

If I forgot to mention anyone, please notify.

Donations

If 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: bc1qwv83ggq8rvv07uk6dv4njs0j3yygj3aax4wg6c

ETH/USDT(e): 0x04C8a749F49aE8a56CB84cF0C99CD9E92eDB17db

XMR: 47F7JAyKP8tMBtzwxpoZsUVB8wzg2VrbtDKBice9FAS1FikbHEXXPof4PAb42CQ5ch8p8Hs4RvJuzPHDtaVSdQzD6ZbA5TZ

if you can offer gpu-time (a100+) - PM.

-
+
- +
-
Prompt
by ushiyama ame, by sub-res, masterpiece, best quality, -1girl, myuri \(spice and wolf\), wolf girl, wolf ears, wolf tail, large tail, silver hair, red eyes, choker, medium breasts, fang, pouch, -black kimono with floral print, thighhighs, hair ribbon, cleavage, obi, holding own tail, posing, :3, claw pose, -outdoors, night, pagoda, lanterns, bamboo forest
Negative prompt
worst quality, low quality, (curvy:0.8)
Seed
1554352247
Size
832x1216
Model
rouwei_07_epsilon
Sampler
Euler Ancestral CFG++
Steps
28
CFG scale
1.2
+
Prompt
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), +<lora:youmuanon_noob-vpred0.1_flawless_ds_v2:1>
Negative prompt
worst quality, bad quality, lowres, jpeg artifacts, bad anatomy, bad hands, multiple views, signature, watermark, censored, covered navel, limited range, high brightness, high contrast,
Seed
4147779862
Size
1024x1024
Model
rouwei_0.6_vpred_rc1
Sampler
Euler a
Steps
28
CFG scale
5
-
Vae
sdxl_vae.safetensors
Hashes
{'vae': '235745af8d', 'model': 'd17fec3a72'}
Version
f1.4.5dev-v1.10.1RC-latest-1503-g13a88f28
Emphasis
No norm
Resources
[{'hash': 'd17fec3a72', 'name': 'rouwei_07_epsilon', 'type': 'model'}]
Model hash
d17fec3a72
Hires steps
22
Hires sampler
Euler CFG++
Hires upscale
1.5
Schedule type
Beta
Hires upscaler
Latent
Mswmsa_enabled
False
Raunet_enabled
False
Adetailer model
face_yolov8n.pt
Adetailer version
24.11.1
Denoising strength
0.65
Adetailer mask blur
4
Adetailer confidence
0.3
Mswmsa_simple_enabled
False
Raunet_simple_enabled
False
Adetailer dilate erode
4
Adetailer inpaint padding
32
Adetailer denoising strength
0.4
Adetailer inpaint only masked
True
-
+
Vae
sdxl-vae-fp16-fix.safetensors
Ensd
31337
Hashes
{'vae': '62c7c729ad', 'model': 'e63d568c63', 'lora:youmuanon_noob-vpred0.1_flawless_ds_v2': '7b7e860a9bae'}
Version
f1.3.6dev2-v1.10.1RC-latest-1453-g0fd79513
Emphasis
No norm
Resources
[{'hash': '7b7e860a9bae', 'name': 'youmuanon_noob-vpred0.1_flawless_ds_v2', 'type': 'lora', 'weight': 1}, {'hash': 'e63d568c63', 'name': 'rouwei_0.6_vpred_rc1', 'type': 'model'}]
Model hash
e63d568c63
Hires steps
18
Hires upscale
1.5
Schedule type
Align Your Steps
Discrete_zsnr
True
Hires upscaler
2x-AnimeSharpV3
Adetailer model
face_yolov8n.pt
Adetailer version
24.9.0
Discrete_sampling
v_prediction
Sigmas_merge_mode
Multiply
Denoising strength
0.5
Sigmas_mult_factor
2
Adetailer mask blur
8
Adetailer confidence
0.3
Sigmas_merge_enabled
True
Adetailer dilate erode
4
Advanced_sampling_mode
Discrete
Adetailer inpaint padding
32
Adetailer negative prompt
{}
Advanced_sampling_enabled
True
Adetailer denoising strength
0.4
Old prompt editing timelines
True
Adetailer inpaint only masked
True
+
- +
-
Prompt
masterpiece, best quality, by pmx, -1girl, provence \(arknights\), wolf ears, wolf tail, large tail, provence \(casual vacation\) \(arknights\), bikini under clothes, cleavage cutout, hair between eyes, hip vent, holding ice cream, holding own tail, long hair, purple hair, short shorts, swimsuit, off-shoulder shirt, yellow eyes, fingerless gloves, -calm atmosphere, epic landscape, -night, dusk, sidelighting, fireflyes, trees, woods, japanese temple, garden, stones, rocks, pagoda, grass, summer, snowflakes, snow on grass,
Negative prompt
worst quality, low quality, nsfw
Seed
594231145
Size
832x1216
Model
rouwei_07_epsilon
Sampler
Euler a
Steps
28
CFG scale
7
+
Prompt
(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
Negative prompt
worst quality, low quality, watermark
Seed
2290774044
Size
832x1216
Model
rouwei_0.6_vpred
Sampler
Euler a
Steps
28
CFG scale
7
-
Vae
sdxl_vae.safetensors
Ensd
31337
Hashes
{'vae': 'a9cb1c1d61', 'model': 'd17fec3a72'}
Version
v1.10.1-72-g021154d8
Emphasis
No norm
Resources
[{'hash': 'd17fec3a72', 'name': 'rouwei_07_epsilon', 'type': 'model'}]
Model hash
d17fec3a72
Hires steps
22
Hires upscale
1.5
Schedule type
Automatic
Hires upscaler
Latent
Adetailer model
face_yolov8n.pt
Adetailer version
24.9.0
Denoising strength
0.6
Adetailer mask blur
4
Adetailer confidence
0.3
Adetailer dilate erode
4
Adetailer inpaint padding
32
Adetailer denoising strength
0.4
Adetailer inpaint only masked
True
-
+
Vae
sdxl_vae.safetensors
Hashes
{'vae': '235745af8d', 'model': '84c8888446'}
Version
f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32
Emphasis
No norm
Resources
[{'hash': '84c8888446', 'name': 'rouwei_0.6_vpred', 'type': 'model'}]
Model hash
84c8888446
Hires steps
18
Hires upscale
1.6
Schedule type
Automatic
Discrete_zsnr
True
Hires upscaler
Latent
Discrete_sampling
v_prediction
Denoising strength
0.6
Advanced_sampling_mode
Discrete
Advanced_sampling_enabled
True
+
- +
-
Prompt
(by muk \(monsieur\):0.8), (by merrytail:0.7), best quality, masterpiece, -1girl, suzuran \(arknights\), fox girl, fox ears, fox tails, kitsune, blonde, hair, hair rings, green eyes, blue hairband, blue t-shirt with text "Cute & Funny", shorts, standing on one leg, balancing on cliff, full body, -Grand Canyon, Colorado River, Arizona, epic landscape
Negative prompt
bad anatomy, bad hands, error, worst quality, low quality
Seed
1806961383
Size
800x1344
Model
rouwei_07_epsilon
Sampler
Euler a
Steps
28
CFG scale
7
+
Prompt
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
Negative prompt
worst quality, low quality, limited range
Seed
1178610355
Size
1152x896
Model
rouwei_0.6_vpred
Sampler
Euler a
Steps
28
CFG scale
7
-
Vae
sdxl_vae.safetensors
Ensd
31337
Hashes
{'vae': 'a9cb1c1d61', 'model': 'd17fec3a72'}
Version
v1.10.1-72-g021154d8
Emphasis
No norm
Resources
[{'hash': 'd17fec3a72', 'name': 'rouwei_07_epsilon', 'type': 'model'}]
Model hash
d17fec3a72
Hires steps
22
Hires upscale
1.5
Schedule type
Automatic
Hires upscaler
4x_foolhardy_Remacri
Adetailer model
face_yolov8n.pt
Adetailer version
24.9.0
Denoising strength
0.5
Adetailer mask blur
4
Adetailer confidence
0.3
Adetailer dilate erode
4
Adetailer inpaint padding
32
Adetailer denoising strength
0.4
Adetailer inpaint only masked
True
-
+
Vae
sdxl_vae.safetensors
Hashes
{'vae': '235745af8d', 'model': '21e24eab73'}
Version
f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32
Emphasis
No norm
Resources
[{'hash': '21e24eab73', 'name': 'rouwei_0.6_vpred', 'type': 'model'}]
Model hash
21e24eab73
Hires upscale
1.5
Schedule type
Automatic
Discrete_zsnr
True
Hires upscaler
Latent
Discrete_sampling
v_prediction
Denoising strength
0.6
Advanced_sampling_mode
Discrete
Advanced_sampling_enabled
True
+
- +
-
Prompt
masterpiece, best quality, by hendra, -1girl, solo, sussurro \(arknights\), fox girl, fox ears, fox tail, grey hair, blue eyes, black thighhighs, short skirt, blue shirt, white jacket, holding a (huge syringe:1.1), glowing eyes, sitting, lap pillow invitation, -pure white background
Negative prompt
worst quality, low quality, watermark
Seed
3010797760
Size
800x1344
Model
rouwei_07_epsilon
Sampler
Euler a
Steps
28
CFG scale
7
+
Prompt
(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
Negative prompt
worst quality, low quality, extra ears
Seed
1556719420
Size
1216x832
Model
rouwei_0.6_vpred
Sampler
Euler a
Steps
20
CFG scale
7
-
Vae
sdxl_vae.safetensors
Ensd
31337
Hashes
{'vae': 'a9cb1c1d61', 'model': 'd17fec3a72'}
Version
v1.10.1-72-g021154d8
Resources
[{'hash': 'd17fec3a72', 'name': 'rouwei_07_epsilon', 'type': 'model'}]
Model hash
d17fec3a72
Hires steps
22
Hires upscale
1.5
Schedule type
Automatic
Hires upscaler
Latent
Adetailer model
face_yolov8n.pt
Adetailer version
24.9.0
Denoising strength
0.6
Adetailer mask blur
4
Adetailer confidence
0.3
Adetailer dilate erode
4
Adetailer inpaint padding
32
Adetailer denoising strength
0.4
Adetailer inpaint only masked
True
+
Vae
sdxl_vae.safetensors
Hashes
{'vae': '235745af8d', 'model': '21e24eab73'}
Version
f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32
Emphasis
No norm
Resources
[{'hash': '21e24eab73', 'name': 'rouwei_0.6_vpred', 'type': 'model'}]
Model hash
21e24eab73
Hires upscale
1.5
Schedule type
Automatic
Discrete_zsnr
True
Hires upscaler
Latent
Discrete_sampling
v_prediction
Denoising strength
0.6
Advanced_sampling_mode
Discrete
Advanced_sampling_enabled
True
- +
-
Prompt
masterpiece, best quality, by wlop, by nat the lich, -1girl, 1boy, wide shot, tate no yuusha no nariagari, outdoors, medieval, dusk, landscape, hug from behind, BREAK -1boy, iwatani naofumi, black hair, green armor, green eyes, fur, holding shield with gem, BREAK -1girl, raphtalia, animal ears, tail, brown hair, purple eyes, white shirt, brown dress, elbow gloves, thigh boots, breastplate, happy
Negative prompt
worst quality, low quality, watermark, nsfw
Seed
1358322656
Size
832x1216
Model
rouwei_07_epsilon
Sampler
Euler a
Steps
28
CFG scale
7
+
Prompt
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
Negative prompt
worst quality, low quality, limited range, monster, 1other, animal, fox
Seed
1376750140
Size
800x1344
Model
rouwei_0.6_vpred
Sampler
Euler a
Steps
28
CFG scale
7
-
Vae
sdxl_vae.safetensors
Ensd
31337
Hashes
{'vae': 'a9cb1c1d61', 'model': 'd17fec3a72'}
Version
v1.10.1-72-g021154d8
Resources
[{'hash': 'd17fec3a72', 'name': 'rouwei_07_epsilon', 'type': 'model'}]
Model hash
d17fec3a72
Hires steps
18
Hires upscale
1.5
Schedule type
Automatic
Hires upscaler
Latent
Adetailer model
face_yolov8n.pt
Adetailer version
24.9.0
Denoising strength
0.6
Adetailer mask blur
4
Adetailer confidence
0.3
Adetailer dilate erode
4
Adetailer inpaint padding
32
Adetailer denoising strength
0.4
Adetailer inpaint only masked
True
+
Vae
sdxl_vae.safetensors
Hashes
{'vae': '235745af8d', 'model': '21e24eab73'}
Version
f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32
Emphasis
No norm
Resources
[{'hash': '21e24eab73', 'name': 'rouwei_0.6_vpred', 'type': 'model'}]
Model hash
21e24eab73
Hires upscale
1.5
Schedule type
Automatic
Discrete_zsnr
True
Hires upscaler
Latent
Discrete_sampling
v_prediction
Denoising strength
0.6
Advanced_sampling_mode
Discrete
Advanced_sampling_enabled
True
- +
-
Prompt
by imazawa, masterpiece, best quality, the devil \(tarot\), -1girl, remilia scarlet, ascot, bat wings, collared shirt, fangs, pointy ears, frills, [blue|gray] hair, jewelry, mob cap, puffy sleeves, floating in air, from below, upside-down, blood, claw pose, sharp fingernails, slit pupils, -purple sky with dense clouds, tarot \(medium\), card \(medium\)
Negative prompt
worst quality, low quality
Seed
4560721
Size
832x1216
Model
rouwei_07_epsilon
Sampler
Euler a
Steps
28
CFG scale
7
+
Prompt
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
Negative prompt
worst quality, low quality
Seed
2168859150
Size
1216x832
Model
rouwei_0.6_vpred
Sampler
Euler a
Steps
20
CFG scale
7
-
Vae
sdxl_vae.safetensors
Ensd
31337
Hashes
{'vae': 'a9cb1c1d61', 'model': 'd17fec3a72'}
Version
v1.10.1-72-g021154d8
Emphasis
No norm
Resources
[{'hash': 'd17fec3a72', 'name': 'rouwei_07_epsilon', 'type': 'model'}]
Model hash
d17fec3a72
Hires steps
22
Hires upscale
1.5
Schedule type
Automatic
Hires upscaler
4x_foolhardy_Remacri
Adetailer model
face_yolov8n.pt
Adetailer version
24.9.0
Denoising strength
0.5
Adetailer mask blur
4
Adetailer confidence
0.3
Adetailer dilate erode
4
Adetailer inpaint padding
32
Adetailer denoising strength
0.4
Adetailer inpaint only masked
True
-
+
Vae
sdxl_vae.safetensors
Hashes
{'vae': '235745af8d', 'model': '84c8888446'}
Version
f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32
Emphasis
No norm
Resources
[{'hash': '84c8888446', 'name': 'rouwei_0.6_vpred', 'type': 'model'}]
Model hash
84c8888446
Hires steps
18
Hires upscale
1.5
Schedule type
Automatic
Discrete_zsnr
True
Hires upscaler
Latent
Discrete_sampling
v_prediction
Denoising strength
0.6
Advanced_sampling_mode
Discrete
Advanced_sampling_enabled
True
+
- +
-
Prompt
by porforever, best quality, masterpiece, -1girl, platinum \(arknights\), horse girl, horse ears, horse tail, long white hair, ponytail, pale skin, crop top, t-shirt with text mega milk, flat chest, :p, happy, midriff, white shorts, hips, posing, lying on sofa, on back, living room, legs apart, cowboy shot, v over eye, -indoors
Negative prompt
worst quality, low quality
Seed
1195269736
Size
832x1216
Model
rouwei_07_epsilon
Sampler
Euler a
Steps
28
CFG scale
7
+
Prompt
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
Negative prompt
worst quality, low quality, watermark
Seed
1114850771
Size
832x1216
Model
rouwei_0.6_vpred
Sampler
Euler a
Steps
28
CFG scale
5
-
Vae
sdxl_vae.safetensors
Ensd
31337
Hashes
{'vae': 'a9cb1c1d61', 'model': 'd17fec3a72'}
Version
v1.10.1-72-g021154d8
Emphasis
No norm
Resources
[{'hash': 'd17fec3a72', 'name': 'rouwei_07_epsilon', 'type': 'model'}]
Model hash
d17fec3a72
Hires steps
22
Hires upscale
1.5
Schedule type
Automatic
Hires upscaler
4x_foolhardy_Remacri
Adetailer model
face_yolov8n.pt
Adetailer version
24.9.0
Denoising strength
0.5
Adetailer mask blur
4
Adetailer confidence
0.3
Adetailer dilate erode
4
Adetailer inpaint padding
32
Adetailer denoising strength
0.4
Adetailer inpaint only masked
True
+
Vae
sdxl_vae.safetensors
Hashes
{'vae': '235745af8d', 'model': '84c8888446'}
Version
f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32
Emphasis
No norm
Resources
[{'hash': '84c8888446', 'name': 'rouwei_0.6_vpred', 'type': 'model'}]
Model hash
84c8888446
Hires steps
18
Hires upscale
1.5
Schedule type
Automatic
Discrete_zsnr
True
Hires upscaler
Latent
Discrete_sampling
v_prediction
Denoising strength
0.6
Advanced_sampling_mode
Discrete
Advanced_sampling_enabled
True
- +
-
Prompt
by deel \(rkeg\), by kaamin \(mariarose753\), best quality, masterpiece, -1girl, sussurro \(arknights\), fox girl, fox ears, fox tail, sussurro \(summer flower\) \(arknights\), black bikini, headphones, alcohol, ice cream, in cup, lying, posing, ice cube, strawberry, full body, wide shot, cocktail glass, cocktail umbrella, -abstract background, gradient background
Negative prompt
worst quality, low quality
Seed
138127413
Size
832x1216
Model
rouwei_07_epsilon
Sampler
Euler a
Steps
28
CFG scale
7
+
Prompt
by akai sashimi, by nat the lich, by foxykuro, by nekoya \(liu\), masterpiece, best quality, +1girl, kal'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
Negative prompt
worst quality, low quality, watermark, dutch angle
Seed
2542436241
Size
832x1216
Model
rouwei_0.6_vpred
Sampler
Euler a
Steps
28
CFG scale
7
-
Vae
sdxl_vae.safetensors
Ensd
31337
Hashes
{'vae': 'a9cb1c1d61', 'model': 'd17fec3a72'}
Version
v1.10.1-72-g021154d8
Emphasis
No norm
Resources
[{'hash': 'd17fec3a72', 'name': 'rouwei_07_epsilon', 'type': 'model'}]
Model hash
d17fec3a72
Hires steps
22
Hires upscale
1.5
Schedule type
Automatic
Hires upscaler
Latent
Adetailer model
face_yolov8n.pt
Adetailer version
24.9.0
Denoising strength
0.6
Adetailer mask blur
4
Adetailer confidence
0.3
Adetailer dilate erode
4
Adetailer inpaint padding
32
Adetailer denoising strength
0.4
Adetailer inpaint only masked
True
+
Vae
sdxl_vae.safetensors
Ensd
31337
Hashes
{'vae': '235745af8d', 'model': '21e24eab73'}
Version
f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32
Emphasis
No norm
Resources
[{'hash': '21e24eab73', 'name': 'rouwei_0.6_vpred', 'type': 'model'}]
Model hash
21e24eab73
Hires steps
18
Hires upscale
1.5
Schedule type
Automatic
Discrete_zsnr
True
Hires upscaler
Latent
Discrete_sampling
v_prediction
Denoising strength
0.6
Advanced_sampling_mode
Discrete
Advanced_sampling_enabled
True
- +
-
Prompt
by meadow \(morphinecaca\), masterpiece, -1girl, nagato \(azur lane\), fox girl, fox ears, fox tail, yellow eyes, black hair, long hair, hair ribbon, hair ornament,multiple tails, kitsune, head tilt, shorts, hoodie, glowing eyes, naughty face, clothes lift, midriff, underboob, posing BREAK -abstract background, (glitch:1.4), vhs, (surreal:1.3), (epic landscape:1.2), underwater, space, black hole, dutch angle, perspective, geometry, shards, partially submerged
Negative prompt
worst quality, low quality
Seed
2521983662
Size
896x1152
Model
rouwei_07_epsilon
Sampler
Euler a
Steps
28
CFG scale
7
+
Prompt
by akai sashimi, by nat the lich, by foxykuro, by nekoya \(liu\), masterpiece, best quality, +1girl, kal'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
Negative prompt
worst quality, low quality, watermark, abstract background
Seed
3956367149
Size
832x1216
Model
rouwei_0.6_vpred
Sampler
Euler a
Steps
20
CFG scale
7
-
Vae
sdxl_vae.safetensors
Ensd
31337
Hashes
{'vae': 'a9cb1c1d61', 'model': 'd17fec3a72'}
Version
v1.10.1-72-g021154d8
Emphasis
No norm
Resources
[{'hash': 'd17fec3a72', 'name': 'rouwei_07_epsilon', 'type': 'model'}]
Model hash
d17fec3a72
Hires steps
22
Hires upscale
1.5
Schedule type
Automatic
Hires upscaler
Latent
Adetailer model
face_yolov8n.pt
Adetailer version
24.9.0
Denoising strength
0.65
Adetailer mask blur
4
Adetailer confidence
0.3
Adetailer dilate erode
4
Adetailer inpaint padding
32
Adetailer denoising strength
0.4
Adetailer inpaint only masked
True
-
+
Vae
sdxl_vae.safetensors
Hashes
{'vae': '235745af8d', 'model': '21e24eab73'}
Version
f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32
Emphasis
No norm
Resources
[{'hash': '21e24eab73', 'name': 'rouwei_0.6_vpred', 'type': 'model'}]
Model hash
21e24eab73
Hires steps
18
Hires upscale
1.6
Schedule type
Automatic
Hires upscaler
Latent
Denoising strength
0.6
+
- +
-
Prompt
by floofsmear, by sub-res, masterpiece, best quality, (text "Dead End":1.2), -1girl, jessica \(arknights\), cat girl, green eyes, dark blue hair, hair ribbon, shorts, hoodie, calm, looking at viewer, glowing eyes, evil smile, sitting in front of table, -annoyed, shaded face, arm behind head, evil smile, holding gun, aiming at viewer, shot, firing, energy beam, (glitch, vhs artifacts, chromatic aberration:1.2), smoke, -indoors, office, wide window with snowing outside, winter
Negative prompt
worst quality, low quality
Seed
3298674343
Size
960x1056
Model
rouwei_07_epsilon
Sampler
Euler a
Steps
28
CFG scale
7
+
Prompt
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
Negative prompt
worst quality, low quality, bad hands, watermark +worst quality, low quality
Seed
2135703813
Size
1376x720
Model
rouwei_0.6_vpred
Sampler
Euler a
Steps
20
CFG scale
7
-
Vae
sdxl_vae.safetensors
Ensd
31337
Hashes
{'vae': 'a9cb1c1d61', 'model': 'd17fec3a72'}
Version
v1.10.1-72-g021154d8
Vec cc b
0.0
Vec cc g
0.0
Vec cc r
0.0
Resources
[{'hash': 'd17fec3a72', 'name': 'rouwei_07_epsilon', 'type': 'model'}]
Model hash
d17fec3a72
Vec cc alt
True
Hires steps
22
Vec cc noise
Straight Abs.
Hires upscale
1.5
Schedule type
Automatic
Hires upscaler
4x_foolhardy_Remacri
Vec cc enabled
True
Vec cc scaling
Flat
Vec cc version
2.2.5
Adetailer model
face_yolov8n.pt
Vec cc contrast
0.2
Vec cc proc ade
False
Vec cc proc hrf
False
Adetailer version
24.9.0
Vec cc brightness
-0.3
Vec cc saturation
1.0
Denoising strength
0.49
Adetailer mask blur
4
Adetailer confidence
0.3
Vec cc seed randomize
False
Adetailer dilate erode
4
Adetailer inpaint padding
32
Adetailer denoising strength
0.4
Adetailer inpaint only masked
True
+
Vae
sdxl_vae.safetensors
Hashes
{'vae': '235745af8d', 'model': '84c8888446'}
Version
f1.2.5dev-v1.10.1RC-latest-1321-g9b35fa32
Emphasis
No norm
Resources
[{'hash': '84c8888446', 'name': 'rouwei_0.6_vpred', 'type': 'model'}]
Model hash
84c8888446
Hires upscale
1.5
Schedule type
Automatic
Hires upscaler
Latent
Denoising strength
0.6
Rescale_cfg_enabled
True
Rescale_cfg_multiplier
0.7
\ No newline at end of file