SDXL_anime_vae

SDXL_anime_vae

• Guide (external site): English | Japanese

This repository provides VAEs (Variational Autoencoders) optimized for SDXL anime illustration models.

SDXL_anime_natural_vae

Test Workflow

<div style="text-align: center; margin-left: auto; margin-right: auto; width: 800px; max-width: 90%;"> <img src="./images/image_difference_checker_sdxl_anime_natural_vae.png" alt="test workflow" width="800" height="400"> </div>

Sample Image

<div style="text-align: center; margin-left: auto; margin-right: auto; width: 400px;"> <img src="./images/sample_image.png" alt="sample_image" width="400" height="400"> </div>

Difference Map

<div style="display: flex; justify-content: center; align-items: center; gap: 2em;"> <div> <img src="./images/SDXL_anime_natural_vae_diff_map.png" alt="SDXL_anime_natural_vae_diff_map" width="400" height="400"> </div> <div> <img src="./images/SDXL_0.9_vae_diff_map.png" alt="SDXL_0.9_vae_diff_map" width="400" height="400"> </div> </div> <p style="text-align:center;">SDXL_anime_natural_vae | SDXL_Base_0.9_vae (Original)</p>

SDXL_anime_natural_vae is a VAE fine-tuned based on SDXL_base_0.9_vae. Compared to the base model, it significantly minimizes fluctuations in MAE and SSIM, prevents green color casting, and produces images that are closer to the original with minimal degradation.

While this model was specifically tuned for anime illustrations, it is also expected to be effective for realistic photo-style illustrations.

Usage

Simply select this model as a replacement for your original VAE.

Precautions

• Recommended Formats: It is highly recommended to use FP32 or BF16.
• FP16 Issues: Because this model inherits the characteristics of the original SDXL_base_0.9_vae, running it in FP16 may result in "black images" (NaN errors).
• Alternative: If you specifically need to perform VAE processing in FP16, we recommend using the sdxl-vae-fp16-fix model instead of this one.

---

SDXL_anime_clear_vae

<div style="display: flex; justify-content: center; align-items: center; gap: 2em;"> <div> <img src="./images/mellow_pencil-XL-v1.0.0-base_clear_upscale_1.png" alt="mellow_pencil-XL-v1.0.0-base_clear_upscale_1" width="400" height="400"> </div> <div> <img src="./images/mellow_pencil-XL-v1.0.0-base_clear_upscale_2.png" alt="mellow_pencil-XL-v1.0.0-base_clear_upscale_2" width="400" height="400"> </div> </div>

Comparison with SDXL_Base_0.9_vae (Original)

<div style="display: flex; justify-content: center; align-items: center; gap: 2em;"> <div> <img src="./images/SDXL_clear_vae_girl.png" alt="SDXL_clear_vae_girl" width="400" height="400"> </div> <div> <img src="./images/SDXL_original_vae_girl.png" alt="SDXL_original_vae_girl" width="400" height="400"> </div> </div> <p style="text-align:center;">SDXL_clear_vae | SDXL_Base_0.9_vae (Original)</p> <div style="display: flex; justify-content: center; align-items: center; gap: 2em;"> <div> <img src="./images/SDXL_clear_vae_bird.png" alt="SDXL_clear_vae_bird" width="400" height="400"> </div> <div> <img src="./images/SDXL_original_vae_bird.png" alt="SDXL_original_vae_bird" width="400" height="400"> </div> </div> <div style="text-align: center; margin-left: auto; margin-right: auto; width: 800px; max-width: 90%;"> <img src="./images/SDXL_clear_vae_compare_bird_close_up.png" alt="SDXL_clear_vae_compare_bird_close_up" width="800" height="400"> </div>

It features clearer rendering compared to the original and is characterized by deeper color expression, primarily in the red spectrum.

However, in areas of high complexity, there is a slight tendency for extremely fine details to be lost or simplified.

---

Base Model

• mellow_pencil-XL-v1.0.0-base_clear
• Animagine-XL_4.0_Opt_clear

---

Tips (External site)

• Color Prompts: English | Japanese
• Color Correction Custom Nodes in ComfyUI: English | Japanese

---

License

• CreativeML Open RAIL++-M