Ming-Image-0.1 VAE

The original VAE of inclusionAI's Ming-Image-0.1, so you can download it from Civitai.

What it is

• The autoencoder of Ming-Image-0.1-Design: 16 latent channels, 127M parameters, bf16, 254 MB

• It works on four channels, RGB plus alpha, where the Qwen Image VAE it shares its layout with works on three

• Not retrained and not sharpened

Works with

• Ming-Image-0.1-Design, whose files are on Civitai here: https://civarchive.com/models/2961185

Install

• Put it in ComfyUI/models/vae/ and select it in the Load VAE node

• Needs ComfyUI with native Ming Image support: the main branch from 24 September 2026, or the next release

About the images

Rendered with Ming-Image-0.1-Design (int8 convrot, 12 steps, cfg 1, 832x1248) and decoded with this file. Raw decodes, no upscaling.

中文：Ming-Image-0.1 原版 VAE（inclusionAI，MIT 许可），未经修改（bf16，254 MB），与 Comfy-Org 发布的文件逐字节一致。16 个潜空间通道，RGBA 四通道。放入 ComfyUI/models/vae/，需要原生支持 Ming 的 ComfyUI。

License

MIT, Copyright (c) 2026 inclusionAI.