Rename to texture-fix-vae-for-qwen-image-2.1 (Qwen license §4(c)): rename ComfyUI file + demo images, update README/NOTICE

.gitattributes

@@ -37,3 +37,5 @@ images/decode-qwen-image-2.1-vae-full.png filter=lfs diff=lfs merge=lfs -text
 images/decode-qwen-image-2.1-vae-texture-fix-full.png filter=lfs diff=lfs merge=lfs -text
 images/decode-qwen-image-2.1-vae-texture-fix.png filter=lfs diff=lfs merge=lfs -text
 images/decode-qwen-image-2.1-vae.png filter=lfs diff=lfs merge=lfs -text
+images/decode-texture-fix-vae-full.png filter=lfs diff=lfs merge=lfs -text
+images/decode-texture-fix-vae.png filter=lfs diff=lfs merge=lfs -text

NOTICE

@@ -1,5 +1,5 @@
 Qwen is licensed under the Qwen RESEARCH LICENSE AGREEMENT, Copyright (c) 2026 Hangzhou Tongyi Laboratory Technology Co., Ltd. All Rights Reserved.
 
-Modified files: diffusion_pytorch_model.safetensors and qwen_image_2.1_vae_texture_fix_bf16.safetensors contain the
+Modified files: diffusion_pytorch_model.safetensors and texture_fix_vae_for_qwen_image_2.1_bf16.safetensors contain the
 Qwen-Image-2.1 VAE (https://huggingface.co/Qwen/Qwen-Image-2.1) with modified (finetuned) decoder weights.
 The modifications were made by madebyollin in 2026. The encoder weights are unchanged.

README.md

@@ -9,11 +9,11 @@ tags:
 - qwen-image
 ---
 
-# Qwen-Image-2.1-VAE-Texture-Fix
+# Texture-Fix VAE for Qwen-Image-2.1
 
-Qwen-Image-2.1-VAE-Texture-Fix is the [Qwen-Image-2.1 VAE](https://huggingface.co/Qwen/Qwen-Image-2.1/tree/main/vae), but finetuned to produce cleaner textures with no checkerboard artifacts.
+The Texture-Fix VAE is the [Qwen-Image-2.1 VAE](https://huggingface.co/Qwen/Qwen-Image-2.1/tree/main/vae), but finetuned to produce cleaner textures with no checkerboard artifacts.
 
-Qwen-Image-2.1-VAE-Texture-Fix's improved decoding is most noticeable in detailed, photo-style images.
+The Texture-Fix VAE's improved decoding is most noticeable in detailed, photo-style images.
 
 <details>
 <summary>Comparison Settings</summary>
@@ -24,17 +24,17 @@ The latents for the VAE comparison image below were generated by Qwen-Image-2.1
 
 </details>
 
-| Qwen-Image-2.1-VAE ([full-res](./images/decode-qwen-image-2.1-vae-full.png)) | 🪄 Qwen-Image-2.1-VAE-Texture-Fix 🪄 ([full-res](./images/decode-qwen-image-2.1-vae-texture-fix-full.png)) |
+| Qwen-Image-2.1-VAE ([full-res](./images/decode-qwen-image-2.1-vae-full.png)) | 🪄 Texture-Fix VAE 🪄 ([full-res](./images/decode-texture-fix-vae-full.png)) |
 | --- | --- |
-| ![](./images/decode-qwen-image-2.1-vae-zoomed-1.png) | ![](./images/decode-qwen-image-2.1-vae-texture-fix-zoomed-1.png) |
-| ![](./images/decode-qwen-image-2.1-vae-zoomed-2.png) | ![](./images/decode-qwen-image-2.1-vae-texture-fix-zoomed-2.png) |
-| ![](./images/decode-qwen-image-2.1-vae.png) | ![](./images/decode-qwen-image-2.1-vae-texture-fix.png) |
+| ![](./images/decode-qwen-image-2.1-vae-zoomed-1.png) | ![](./images/decode-texture-fix-vae-zoomed-1.png) |
+| ![](./images/decode-qwen-image-2.1-vae-zoomed-2.png) | ![](./images/decode-texture-fix-vae-zoomed-2.png) |
+| ![](./images/decode-qwen-image-2.1-vae.png) | ![](./images/decode-texture-fix-vae.png) |
 
 # Usage
 
 **ComfyUI**
 
-Download [`qwen_image_2.1_vae_texture_fix_bf16.safetensors`](./qwen_image_2.1_vae_texture_fix_bf16.safetensors) into `ComfyUI/models/vae/` and select it in the `Load VAE` node, in place of `qwen_image_2.1_vae_bf16.safetensors`.
+Download [`texture_fix_vae_for_qwen_image_2.1_bf16.safetensors`](./texture_fix_vae_for_qwen_image_2.1_bf16.safetensors) into `ComfyUI/models/vae/` and select it in the `Load VAE` node, in place of `qwen_image_2.1_vae_bf16.safetensors`.
 
 **🧨 Diffusers**
 
@@ -42,13 +42,13 @@ Download [`qwen_image_2.1_vae_texture_fix_bf16.safetensors`](./qwen_image_2.1_va
 import torch
 from diffusers import QwenImage21Pipeline, AutoencoderKLQwenImage21
 
-vae = AutoencoderKLQwenImage21.from_pretrained("madebyollin/qwen-image-2.1-vae-texture-fix", torch_dtype=torch.bfloat16)
+vae = AutoencoderKLQwenImage21.from_pretrained("madebyollin/texture-fix-vae-for-qwen-image-2.1", torch_dtype=torch.bfloat16)
 pipe = QwenImage21Pipeline.from_pretrained("Qwen/Qwen-Image-2.1", vae=vae, torch_dtype=torch.bfloat16).to("cuda")
 ```
 
 # Mechanism
 
-Qwen-Image-2.1-VAE-Texture-Fix was created by finetuning the Qwen-Image-2.1 VAE decoder for ~5000 steps at learning rate 3e-5, with only the two highest-resolution decoder stages and output head unfrozen (7.5M trainable parameters), using the recipe developed for [TAESD](https://github.com/madebyollin/taesd).
+The Texture-Fix VAE was created by finetuning the Qwen-Image-2.1 VAE decoder for ~5000 steps at learning rate 3e-5, with only the two highest-resolution decoder stages and output head unfrozen (7.5M trainable parameters), using the recipe developed for [TAESD](https://github.com/madebyollin/taesd).
 
 The TAESD recipe, like most image autoencoder training recipes, uses a mix of PSNR-focused (MSE/MAE), LPIPS, and adversarial (GAN) loss terms.
 Whenever precise details can't be reconstructed, MSE/MAE loss encourages blurring, LPIPS loss encourages blurring+checkerboarding (among other artifacts), and adversarial loss encourages generating sharp/plausible (but fake) detail without obvious artifacts.
@@ -60,9 +60,9 @@ I suspect the original Qwen-Image-2.1-VAE was trained without a working adversar
 
 # Metrics
 
-Qwen-Image-2.1-VAE-Texture-Fix makes perceptual quality metrics (rFID) better and reconstruction accuracy metrics (LPIPS/PSNR) slightly worse.
+The Texture-Fix VAE makes perceptual quality metrics (rFID) better and reconstruction accuracy metrics (LPIPS/PSNR) slightly worse.
 
-| Metric | Qwen-Image-2.1-VAE | Qwen-Image-2.1-VAE-Texture-Fix |
+| Metric | Qwen-Image-2.1-VAE | Texture-Fix VAE |
 | --- | --- | --- |
 | rFID ↓ (COCO val2017, 5000 images @ 256²) | 3.37 | **2.08** |
 | PSNR ↑ (COCO val2017 @ 256²) | **33.30** | 32.86 |