Upload folder using huggingface_hub

.gitattributes

@@ -294,3 +294,10 @@ ComfyUI/models/sams/sam_vit_b_01ec64.pth filter=lfs diff=lfs merge=lfs -text
 ComfyUI/models/sams/sams/sam_vit_b_01ec64.pth filter=lfs diff=lfs merge=lfs -text
 ComfyUI/models/text_encoders/ernie-image-prompt-enhancer.webp filter=lfs diff=lfs merge=lfs -text
 ComfyUI/models/text_encoders/ltx-2.3_text_projection_bf16.mp4 filter=lfs diff=lfs merge=lfs -text
+ComfyUI/models/vae/LTX23_audio_vae_bf16.mp4 filter=lfs diff=lfs merge=lfs -text
+ComfyUI/models/vae/LTX23_video_vae_bf16.mp4 filter=lfs diff=lfs merge=lfs -text
+ComfyUI/models/vae/UltraFlux-v1-vaer.webp filter=lfs diff=lfs merge=lfs -text
+ComfyUI/models/vae/flux2-dev-vae.safetensors filter=lfs diff=lfs merge=lfs -text
+ComfyUI/models/vae/flux2-vae.safetensors filter=lfs diff=lfs merge=lfs -text
+ComfyUI/models/vae/taeltx2_3.safetensors filter=lfs diff=lfs merge=lfs -text
+ComfyUI/models/vae/z-img-ae.preview.jpeg filter=lfs diff=lfs merge=lfs -text

ComfyUI/models/vae/LTX23_audio_vae_bf16.md

@@ -0,0 +1,47 @@
+---
+author: MrReclusive666
+baseModel: LTXV2
+hashes:
+  AutoV1: 1324F6CB
+  AutoV2: 5BC10FA4AD
+  AutoV3: 28CEB51ED018
+  BLAKE3: 3215B775195E9306B7B704970D9B5D15AB40CA42AC5AA861574E83ABEA130684
+  CRC32: EA02D754
+  SHA256: 5BC10FA4ADECF99DDA132D916E23048CBD56797702C5FA50EB5D2079048A38C3
+metadata:
+  format: SafeTensor
+  fp: bf16
+  size: full
+modelPage: https://civitai.com/models/2445970?modelVersionId=2750820
+preview:
+- https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/803c5d4c-b254-47fd-b6e9-0f3b113f79fa/original=true/123401038.mp4
+website: Civitai
+---
+
+# Trigger Words
+
+No trigger words
+
+# About this version
+
+Audio vae for TFO model
+
+# LTX2.3 FP4
+
+**LTX2.3 FP4 Models**
+=====================
+
+### Because no one uploads FP4 here. and they should (fp4 works on 3xxx and above for vram savings.)
+
+LTX2.3 Distilled FP4ME - Distilled FP4 Mixed Extreme - 14.1GB\*\*  
+LTX2.3 Full Dev FP4ME - Full Dev FP4 Mixed Extreme - 14.1GB\*\*  
+LTX2.3 Official NVFP4 - From Lightricks - Updated 3/17/2026 - 21GB  
+LTX2.3 Dev NVFP4 - from [Hippotes /](https://huggingface.co/Hippotes) [LTX-2.3-various-formats](https://huggingface.co/Hippotes/LTX-2.3-various-formats) - 18.15GB  
+LTX2.3 Dev FP4 STFO - from Kijai's Transformers only Scaled FP8 - 16.63GB\*\*  
+LTX2.3 Distilled FP4 - My first FP4 - Transformers Only - 18.15GB\*\*  
+  
+\*\* Are transformers only, or I broke vae/text projection :) needs seperate Vae and Text projection downloads.
+
+I have a fix for lora issue on FP4ME models. Comfy Bathroom, custom lora loader with presets as well as custom config. <https://huggingface.co/MrReclusive/LTX-2.3-FP4>
+
+Diffusion model loader/unet loader seems to get confused with these, use the checkpoint loader even though it has no clip/vae

ComfyUI/models/vae/LTX23_audio_vae_bf16.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd6691da822cfe39fde58eeb62b07f51bfaf215884311d4ecbc8d8b77b23d40a
+size 5073813

ComfyUI/models/vae/LTX23_video_vae_bf16.md

@@ -0,0 +1,47 @@
+---
+author: MrReclusive666
+baseModel: LTXV 2.3
+hashes:
+  AutoV1: '11752536'
+  AutoV2: 01EA62D09B
+  AutoV3: 87C20C045586
+  BLAKE3: 9AFA56159C4EF857994FC98C7E4500A20980852600774F5A69782A4E18369372
+  CRC32: 70441E32
+  SHA256: 01EA62D09BC139F95C5DEE7B5C062AD6A3E6CD8BE910A1983AC02E7EB5B8EE3B
+metadata:
+  format: SafeTensor
+  fp: bf16
+  size: full
+modelPage: https://civitai.com/models/2445970?modelVersionId=2750791
+preview:
+- https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/25f9d8e3-0d8f-4505-b8d8-135a6adc9db6/original=true/123400380.mp4
+website: Civitai
+---
+
+# Trigger Words
+
+No trigger words
+
+# About this version
+
+Vae's For TFO model
+
+# LTX2.3 FP4
+
+**LTX2.3 FP4 Models**
+=====================
+
+### Because no one uploads FP4 here. and they should (fp4 works on 3xxx and above for vram savings.)
+
+LTX2.3 Distilled FP4ME - Distilled FP4 Mixed Extreme - 14.1GB\*\*  
+LTX2.3 Full Dev FP4ME - Full Dev FP4 Mixed Extreme - 14.1GB\*\*  
+LTX2.3 Official NVFP4 - From Lightricks - Updated 3/17/2026 - 21GB  
+LTX2.3 Dev NVFP4 - from [Hippotes /](https://huggingface.co/Hippotes) [LTX-2.3-various-formats](https://huggingface.co/Hippotes/LTX-2.3-various-formats) - 18.15GB  
+LTX2.3 Dev FP4 STFO - from Kijai's Transformers only Scaled FP8 - 16.63GB\*\*  
+LTX2.3 Distilled FP4 - My first FP4 - Transformers Only - 18.15GB\*\*  
+  
+\*\* Are transformers only, or I broke vae/text projection :) needs seperate Vae and Text projection downloads.
+
+I have a fix for lora issue on FP4ME models. Comfy Bathroom, custom lora loader with presets as well as custom config. <https://huggingface.co/MrReclusive/LTX-2.3-FP4>
+
+Diffusion model loader/unet loader seems to get confused with these, use the checkpoint loader even though it has no clip/vae

ComfyUI/models/vae/LTX23_video_vae_bf16.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f21be876b6b6628adad9958f46ce0c3df7db759a2130132a92aafb199a863c04
+size 2767115

ComfyUI/models/vae/UltraFlux-v1-vaer.md

@@ -0,0 +1,48 @@
+---
+author: jorkingtoncityshallwe
+baseModel: ZImageTurbo
+hashes:
+  AutoV1: 6778205B
+  AutoV2: 2BF9AD6856
+  AutoV3: 47B06942F183
+  BLAKE3: 1C69412A06338E15D285AA8835606F1768C0906BC13A069CF4CC3E46607BFB65
+  CRC32: '52402561'
+  SHA256: 2BF9AD685686B480B03651A8D8595951E4A5578016B8EAD4AF5E22D3DC9B3409
+metadata:
+  format: SafeTensor
+modelPage: https://civitai.com/models/2231253?modelVersionId=2511862
+preview:
+- https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/90109f71-9bf8-4779-857b-42412f5f2861/original=true/114170661.jpeg
+- https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/49e170c3-b581-4b07-987c-71d9a3fa595f/original=true/114170667.jpeg
+- https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/dcba0042-a0ec-4ad9-9076-f13deddc116d/original=true/114170663.jpeg
+- https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/6b72df2c-7fb7-4b89-ae99-020337a833f9/original=true/114170664.jpeg
+- https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/c9b2d8a8-16b9-46dd-93a8-2b4d02c147b1/original=true/114170662.jpeg
+- https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/e9b29cab-e4c6-4c07-9d72-54736e3051e8/original=true/114170668.jpeg
+- https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/b6399540-3f4b-4bd9-b3fc-dddc3eaa628c/original=true/114170666.jpeg
+website: Civitai
+---
+
+# Trigger Words
+
+No trigger words
+
+# About this version
+
+This VAE works with all flux and zimage models
+
+# UltraFlux VAE | improved quality for flux and zimage
+
+Detailed comparison between standard VAE and the UltraFlux VAE: <https://imgsli.com/NDM1MTI2>
+=============================================================================================
+
+**This vae is not made by me, I am simply re-uploading the vae from this awesome project** <https://huggingface.co/Owen777/UltraFlux-v1>**!!!**
+
+From my testing it improves the sharpness of any model using the flux vae (flux based models, Zimage, etc).
+
+In rare cases it might look a little over sharpened but from my short testing it looked pretty good.
+
+Links to the project from which this vae originates from:  
+<https://huggingface.co/Owen777/UltraFlux-v1>  
+<https://github.com/W2GenAI-Lab/UltraFlux/tree/main>
+
+I should have clarified from the start, but basically as [**vladulidlo**](https://civitai.com/user/vladulidlo) has mentioned it's best to use this VAE on a second pass after upscaling. In my opinion it is still worth using in simple 1 pass workflows too (which is why I didn't bother overcomplicating stuff) but if you pixel peep and the oversharpening bothers you then use a 2 stage generation workflow for better results.

ComfyUI/models/vae/ae.md

@@ -0,0 +1,29 @@
+---
+author: Ragnar890
+baseModel: Flux.1 D
+hashes:
+  AutoV1: EB6DC28C
+  AutoV2: AFC8E28272
+  AutoV3: DDEC9C299F56
+  BLAKE3: CE21CB76364AA6E2421311CF4A4B5EB052A76C4F1CD207B50703D8978198A068
+  CRC32: 626AAEF6
+  SHA256: AFC8E28272CD15DB3919BACDB6918CE9C1ED22E96CB12C4D5ED0FBA823529E38
+metadata:
+  format: SafeTensor
+modelPage: https://civitai.com/models/636193?modelVersionId=711305
+preview:
+- https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/83da9b79-608e-462e-a286-8edc3beba968/original=true/23628003.jpeg
+website: Civitai
+---
+
+# Trigger Words
+
+No trigger words
+
+# About this version
+
+ae.sft
+
+# Flux vae.sft
+
+Flux VAe.sft

ComfyUI/models/vae/flux2-dev-vae.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d64f3a68e1cc4f9f4e29b6e0da38a0204fe9a49f2d4053f0ec1fa1ca02f9c4b5
-size 336213556
+oid sha256:868fe7b343cc8f3a19dbcfcafbc3d5f888802be3f89bd81b65b3621a066ce8f3
+size 336211292

ComfyUI/models/vae/flux2-vae.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d64f3a68e1cc4f9f4e29b6e0da38a0204fe9a49f2d4053f0ec1fa1ca02f9c4b5
-size 336213556
+oid sha256:868fe7b343cc8f3a19dbcfcafbc3d5f888802be3f89bd81b65b3621a066ce8f3
+size 336211292

ComfyUI/models/vae/qwen-image/qwen_image_vae.md

@@ -0,0 +1,33 @@
+---
+author: Hikarias
+baseModel: Other
+hashes:
+  AutoV1: 05DC78C1
+  AutoV2: A70580F021
+  AutoV3: 4F452011EA09
+  BLAKE3: 1018B941788940C4DFC449D5D749950CAE58C9A7F6A794B00A2F8321ABC885D1
+  CRC32: 15C82C0A
+  SHA256: A70580F0213E67967EE9C95F05BB400E8FB08307E017A924BF3441223E023D1F
+metadata:
+  format: SafeTensor
+  fp: bf16
+  size: full
+modelPage: https://civitai.com/models/1842123?modelVersionId=2089517
+preview:
+- https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/944f5833-5594-43ff-9911-49ca183e0f69/original=true/92798468.jpeg
+website: Civitai
+---
+
+# Trigger Words
+
+No trigger words
+
+# About this version
+
+No description about this version
+
+# Qwen-Image - GGUF
+
+This is a direct GGUF conversion of [**Qwen/Qwen-Image**](https://huggingface.co/Qwen/Qwen-Image).
+
+The model files can be used in [**ComfyUI**](https://github.com/comfyanonymous/ComfyUI/) with the [**ComfyUI-GGUF**](https://github.com/city96/ComfyUI-GGUF) custom node. Place the required model(s) in the following folders:

ComfyUI/models/vae/taeltx2_3.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6e4cc0469134213d0101a46877ea2bce1dc7cf06ff5f5aefb9e4076c03542f7b
+oid sha256:f0773b4e3e57318e6aa4dd4a35e1d16213a5f160fbc0376163f06888bbcbe246
 size 23531296

ComfyUI/models/vae/z-img-ae.cminfo.json

@@ -0,0 +1,52 @@
+{
+  "ModelId": 2169712,
+  "ModelName": "Z-Image Turbo - Quantized for low VRAM",
+  "ModelDescription": "<p><strong>Z-Image Turbo</strong> is a distilled version of Z-Image, a 6B image model based on the Lumina architecture, developed by the Tongyi Lab team at Alibaba Group. Source: <a target=\"_blank\" rel=\"ugc\" href=\"https://huggingface.co/Tongyi-MAI/Z-Image-Turbo\">https://huggingface.co/Tongyi-MAI/Z-Image-Turbo</a></p><p>I've uploaded quantized versions from bf16 to fp8, meaning the weights had their precision - and consequently their size - halved for a substantial performance boost while keeping most of the quality. Inference time should be similar to regular \"undistilled\" SDXL, with better prompt adherence and resolution/details. Ideal for weak(er) PCs.</p><h1 id=\"features\">Features</h1><ul><li><p><strong>Lightweight</strong>: the Turbo version was trained at low steps (5-15), and the fp8 quantization is roughly 6 GB in size, making it accessible even to low-end GPUs.</p></li><li><p><strong>Uncensored</strong>: many concepts censored by other models (&lt;cough&gt; Flux &lt;cough&gt;) are doable out of the box.</p></li><li><p><strong>Good prompt adherence</strong>: comparable to Flux.1 Dev's, thanks to its powerful text encoder Qwen 3 4B.</p></li><li><p><strong>Text rendering</strong>: comparable to Flux.1 Dev's, some say it's even better despite being much smaller (probably not as good as Qwen Image's though).</p></li><li><p><strong>Style flexibility</strong>: photorealistic images are its biggest strength, but it can do anime, oil painting, pixel art, low poly, comics, watercolor, vector art / flat design, comic book, sketch, pop art, infographic, etc.</p></li><li><p><strong>High resolution</strong>: capable of generating up to 4MP resolution natively (i.e. before upscale) while maintaining coherence.</p></li></ul><h1 id=\"dependencies\">Dependencies</h1><ul><li><p>Download <strong>Qwen 3 4B</strong> to your <code>text_encoders</code> directory: <a target=\"_blank\" rel=\"ugc\" href=\"https://civitai.com/models/2169712?modelVersionId=2474529\">https://civitai.com/models/2169712?modelVersionId=2474529</a></p><ul><li><p>Alternatives:</p><ul><li><p>bf16:</p><ul><li><p><a target=\"_blank\" rel=\"ugc\" href=\"https://civitai.com/models/2168935?modelVersionId=2442540\">Civitai</a></p></li><li><p><a target=\"_blank\" rel=\"ugc\" href=\"https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/text_encoders/qwen_3_4b.safetensors\">ComfyUI</a></p></li></ul></li><li><p>GGUF (for even smaller file):</p><ul><li><p><a target=\"_blank\" rel=\"ugc\" href=\"https://huggingface.co/Qwen/Qwen3-4B-GGUF/blob/main/Qwen3-4B-Q8_0.gguf\">Q8_0</a></p></li></ul></li></ul></li></ul></li><li><p>Download <strong>Flux VAE</strong> to your <code>vae</code> directory: <a target=\"_blank\" rel=\"ugc\" href=\"https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors\">https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors</a></p><ul><li><p>Mirror: <a target=\"_blank\" rel=\"ugc\" href=\"https://civitai.com/models/2168935?modelVersionId=2442479\">Civitai</a></p></li><li><p>TAEF1 (for even smaller file): <a target=\"_blank\" rel=\"ugc\" href=\"https://github.com/madebyollin/taesd/blob/main/taef1_decoder.pth\">decoder</a> and <a target=\"_blank\" rel=\"ugc\" href=\"https://github.com/madebyollin/taesd/blob/main/taef1_encoder.pth\">encoder</a> (download to your <code>vae_approx</code> folder)</p></li></ul></li></ul><h1 id=\"instructions\">Instructions</h1><p>Workflow and metadata are available in the showcase images.</p><ul><li><p><strong>Steps</strong>: 5 - 15 (6 - 11 is the sweet spot)</p></li><li><p><strong>CFG</strong>: 1.0. This will ignore negative prompts, so no need for them.</p></li><li><p><strong>Sampler/scheduler</strong>: depends on the art style. Here are my findings so far:</p><ul><li><p><strong>Photorealistic:</strong></p><ul><li><p>Favourite combination for the base image: <code>euler</code> + <code>beta</code>, <code>simple</code> or <code>bong_tangent</code> (from <a target=\"_blank\" rel=\"ugc\" href=\"https://github.com/ClownsharkBatwing/RES4LYF/\">RES4LYF</a>) - fast and good even at low (5) steps.</p></li><li><p>Most multistep samplers (e.g.: <code>res_2s</code>, <code>res_2m</code>, <code>dpmpp_2m_sde</code> etc) are great, but some will be 40% slower at same steps. They might work better with a scheduler like <code>sgm_uniform</code>.</p></li><li><p>Almost any sampler will work fine - <code>sa_solver</code>, <code>seeds_2</code>, <code>er_sde</code>, <code>gradient_estimation</code>.</p></li><li><p>Some samplers and schedulers add too much texture, you can adjust it by increasing the shift (e.g.: set shift 7 in ComfyUI's <code>ModelSamplingAuraFlow</code> node).</p></li><li><p>Some require more steps (e.g.: <code>karras</code>)</p></li></ul></li><li><p><strong>Illustrations (e.g.: anime):</strong></p><ul><li><p><code>res_2m</code> or <code>rk_beta</code> produce sharper and more colourful results.</p></li></ul></li><li><p><strong>Other styles:</strong></p><ul><li><p>I'm still experimenting. Use <code>euler</code> (or <code>res_2m</code>) + <code>simple</code> just to be safe for now.</p></li></ul></li></ul></li><li><p><strong>Resolution</strong>: up to 4MP native. Avoid going higher than 2048. When in doubt, use same as SDXL, Flux.1, Qwen Image, etc (it works even as low as 512px, like SD 1.5 times). Some examples:</p><ul><li><p>896 x 1152</p></li><li><p>1024 x 1024</p></li><li><p>1216 x 832</p></li><li><p>1440 x 1440</p></li><li><p>1024 x 1536</p></li><li><p>2048 x 2048</p></li></ul></li><li><p><strong>Upscale</strong> and/or <strong>detailers</strong> are recommended to fix smaller details like eyes, teeth, hair. See my workflow embedded in the main cover image.</p><ul><li><p>If going over 2048px in either side, I recommend the tiled upscale method i.e. using UltimateSD Upscale at low denoise (&lt;= 0.3).</p></li><li><p>Otherwise, I recommend your 2nd pass KSampler to either have a low denoise (&lt; 0.5) or to start the sampling at a later step (e.g.: from 5 to 9 steps).</p></li><li><p>Either way, I recommend setting the shift to 7 to avoid noisy textures in your results. Keep in mind that some schedulers (e.g.: <code>bong_tangent</code>) may override the shift with its own.</p></li><li><p>At this stage, you may use even samplers that didn't work well in the initial generation. For most cases, I like the <code>res_2m</code> + <code>simple</code> combination.</p></li></ul></li><li><p><strong>Prompting</strong>: officially they say long and detailed prompts in natural language works best, but I tested with comma-separated keywords/tags, JSON, whatever... either should work fine. Keep it in English or Mandarin for more accurate results.</p></li></ul><h1 id=\"faq\">FAQ</h1><ul><li><p><strong>Is the model uncensored?</strong></p><ul><li><p>Yes, it might just not be well trained on the specific concept you're after. Try it yourself.</p></li></ul></li><li><p><strong>Why do I get too much texture or artifacts after upscaling?</strong></p><ul><li><p>See instructions about upscaling above.</p></li></ul></li><li><p><strong>Does it run on my PC?</strong></p><ul><li><p>If you can run SDXL, chances are you can run Z-Image Turbo fp8. If not, might be a good time to purchase more RAM or VRAM.</p></li><li><p>All my images were generated on a laptop with 32GB RAM, RTX3080 Mobile 8GB VRAM.</p></li></ul></li><li><p><strong>How can I get more variation across seeds?</strong></p><ul><li><p>Start at late step (e.g.: from 3 til 11); or</p></li><li><p>Give clear instructions in prompt, something like <code>give me a random variation of the following image: &lt;your prompt&gt;</code>)</p></li></ul></li><li><p><strong>I'm getting an error on ComfyUI, how to fix it?</strong></p><ul><li><p>Make sure your ComfyUI has been updated to the latest version. Otherwise, feel free to post a comment with the error message so the community can help.</p></li></ul></li><li><p><strong>Is the license permissive?</strong></p><ul><li><p>It's Apache 2.0, so quite permissive.</p></li></ul></li></ul><p></p>",
+  "CreatorUsername": "SimplesmenteIA",
+  "Nsfw": false,
+  "Poi": false,
+  "AllowNoCredit": true,
+  "AllowCommercialUse": "{Image,RentCivit,Rent,Sell}",
+  "AllowDerivatives": true,
+  "AllowDifferentLicense": true,
+  "Tags": [
+    "base model"
+  ],
+  "ModelType": "Checkpoint",
+  "VersionId": 2445746,
+  "VersionName": "fp8_scaled_e4m3fn_KJ",
+  "VersionDescription": "<p>This is a fp8_e4m3fn <strong>scaled</strong> quantization, in which the dynamic range of the weights are supposed to be more optimal, resulting in slightly better quality than normal fp8. </p><p>Almost same size as SDXL, while some say it could be as good as (or even better than) heavier models such as Flux.1 and Qwen Image.</p><p>Credits to <a rel=\"ugc\" href=\"https://huggingface.co/Kijai/Z-Image_comfy_fp8_scaled\">Kijai on huggingface</a>.</p>",
+  "BaseModel": "ZImageTurbo",
+  "BaseModelType": "Standard",
+  "EarlyAccessDeadline": null,
+  "VersionPublishedAt": "2025-11-28T01:47:59.404Z",
+  "VersionUpdatedAt": "2025-12-06T23:55:43.878Z",
+  "VersionStatus": "Published",
+  "IsPrimaryFile": false,
+  "PrimaryFileId": 625974,
+  "PrimaryFileName": "ae.sft",
+  "FileMetadata": {
+    "fp": null,
+    "size": null,
+    "format": "SafeTensor"
+  },
+  "ImportedAt": "2025-12-14T04:03:39.088015+00:00",
+  "Hashes": {
+    "AutoV1": "EB6DC28C",
+    "AutoV2": "AFC8E28272",
+    "SHA256": "AFC8E28272CD15DB3919BACDB6918CE9C1ED22E96CB12C4D5ED0FBA823529E38",
+    "CRC32": "626AAEF6",
+    "BLAKE3": "CE21CB76364AA6E2421311CF4A4B5EB052A76C4F1CD207B50703D8978198A068",
+    "AutoV3": "DDEC9C299F56"
+  },
+  "TrainedWords": [],
+  "Stats": {
+    "downloadCount": 6328,
+    "rating": 0,
+    "ratingCount": 0,
+    "favoriteCount": 0,
+    "commentCount": 48,
+    "thumbsUpCount": 438
+  },
+  "DownloadUrlUsed": "https://civitai.com/api/download/models/2445746?type=VAE&format=SafeTensor"
+}