# MOSS-Image Inpainting revisited — retained record

> Retained reddit record from reddit:comfyui.

Canonical: https://highsignal.app/data/records/2116fdf2ac363f25

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:comfyui)  published 2026-11-07 18:08 ingested 2026-11-08 02:39

[original source ↗](https://www.reddit.com/r/comfyui/comments/1xqctax/mossimage_inpainting_revisited/)

## retained text



```
I [previously tested MOSS Image for inpainting](https://www.reddit.com/r/comfyui/comments/1xe4l3b/mossimage_as_an_inpainting_model_not_sure/) in a fairly unfair test, based on a previous poster's enthusiasm with the base model for this purpose. I wasn't impressed, but there was actually very little reason to think it would be particularly good for inpainting compared to dedicated models like Qwen Image Edit 2.1 or Flux.2 Klein 9B.

But having tried the new ComfyUI image-edit workflows with the Image Edit LoRA, I have to say this is an entirely different beast now. The Edit LoRA does make the model able to do a lot of the stuff the other models do. The workflow is just the MOSS-Image: Image Edit LoRA workflow, adapted for inpainting. I'll attach them in a comment.
```



document fetched 2026-11-08 00:33 · canonical https://reddit.com/r/comfyui/comments/1xqctax/mossimage_inpainting_revisited
