MOSSquid

MOSSquid is a merged MOSS-Image checkpoint for text-to-image generation in ComfyUI. The diffusion weights are available in BF16 and INT8 ConvRot. The INT8 file is about 47% smaller on disk.

File Format Size
MOSSquid_bf16.safetensors BF16 10.16 GB
MOSSquid_int8_convrot.safetensors INT8 ConvRot 5.41 GB

Use in ComfyUI

  1. Put either checkpoint in ComfyUI/models/diffusion_models/.
  2. Get a compatible Qwen3.5 text encoder, FLUX.2 VAE, and the text-to-image workflow from Comfy-Org/MOSS-Image.
  3. Select MOSSquid in the workflow's diffusion model loader. The INT8 variant requires ConvRot support.

日本語

MOSSquidは、ComfyUIで使えるMOSS-Image系の画像生成マージモデルです。BF16版と、ファイルサイズを約47%抑えたINT8 ConvRot版を用意しています。

使いたい方のファイルを ComfyUI/models/diffusion_models/ に配置し、MOSS-ImageのText-to-Imageワークフローで選択してください。テキストエンコーダーとVAEは別途必要です。INT8版を使う場合は、ConvRotに対応したComfyUIを使用してください。

Examples / 生成例

Japanese dialogue in a dojo / 道場と日本語の吹き出し Elf maid in a manor / 洋館のエルフメイド
Anime scene in a dojo with a Japanese speech bubble Full-body illustration of an elf maid in a manor

Credits

Thanks to the OpenMOSS Team for MOSS-Image and to the contributors behind MOEMOSS-Base.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support