Downloads · 30 days
0
Rikunarita-ORG/Anima-illustron-Qwen-Image-VAE
Anima-illustron-Qwen-Image-VAE is a machine learning model from Rikunarita-ORG. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
Downloads · 30 days
0
Access
Public
Updated Oct 23, 2026
Repo size
1 GB
Likes
4
Public
Click a slice to open those files.
.safetensors254 MB · 100%
From the Hugging Face model README
An Optimized VAE Exclusively for Anime Illustration Generation
<div style="display: flex; gap: 10px; justify-content: center; align-items: center; flex-wrap: wrap;">
This is a custom VAE designed specifically for Qwen-Image, optimized for anime illustration generation.
By merging two VAEs with distinct characteristics, it successfully combines the "vibrant colors" of Qwen-Image-HDR-VAE with the "noise-free, natural line art" of Qwen-Image-Clear-VAE.
It is specifically optimized for use with models specialized in anime illustration generation (2D/2.5D).
| Base Model | Pros | Cons |
|---|---|---|
| Qwen-Image-HDR-VAE | High dynamic range delivers vibrant colors (especially enhanced blues). Ideal for anime illustrations. Also offers high resolution resilience. | Excessive edge sharpness causes ringing artifacts, resulting in an unnatural appearance. |
| Qwen-Image-Clear-VAE | Resistant to unwanted artifacts. Produces natural lines with a hand-drawn feel. | Colors can appear flat and sometimes washed out or dull. |
| Characteristic | Base Model | Result in This VAE |
|---|---|---|
| Color Vibrancy | HDR-VAE (30%) | Vibrant colors (enhanced blue tones) and expanded dynamic range |
| Natural Line Art / Noise Resistance | Clear-VAE (70%) | Retains hand-drawn texture while suppressing excessive sharpness / Eliminates fine graininess for smooth gradients |
We have fully disclosed the merge procedure and testing methodology for this model.
By importing these workflows into ComfyUI, you can reproduce the exact same merge under identical conditions.
Quality is not guaranteed for photorealistic (3D) image generation models.
Introducing Anima-illustron-Qwen-Image-VAE: A Dedicated Anime VAE at the Golden 7:3 Ratio
This model is released under the Apache 2.0 License, in accordance with the license of the base models.
We welcome artworks generated using this model, as well as suggestions for improvement.
Feel free to reach out via the 👏 Community tab!
This VAE was created as a companion to "Anima-illustron," an anime-specialized diffusion model currently under development.
"Anima-illustron" itself is developed by merging multiple models, and this VAE has been optimized to maximize its potential.
The main model (Anima-illustron) will be released soon!
A link will be shared on this page once it becomes available.