← sources

Update: ComfyUI-Moss-Image-Edit-Native now supports the HQ edit LoRA, with tested default settings

source: reddit (reddit:StableDiffusion)published 2026-10-24 22:41ingested 2026-10-25 02:38

retained text

An update to the native MOSS-Image edit nodes I posted before:

**HQ LoRA support:** [akhaliq/MOSS-Image-Edit-HQ](https://huggingface.co/akhaliq/MOSS-Image-Edit-HQ) works alongside [chinoll](https://huggingface.co/chinoll/Moss-Image-Edit)’s. Pick it with edit\_format; each LoRA gets the layout it was trained with.

**Tested defaults:** I ran 300+ test edits on both LoRAs to find what works. LoRA strength 1.0 (lower strengths redraw the scene, and two-image edits just return image 2), CFG 5 for chinoll and 3 for HQ, skip layer guidance on layer 22, and the SFT checkpoint rather than Base.

**Automatic sizing:** each image goes in as a \~1 MP bucket of its own aspect ratio, and the output follows image 1.

**Reference cache:** the reference images are computed once per run instead of every step (kv\_cache, auto by default).

New example workflows for both LoRAs, with one and two images.

If you used the older workflows (strength 0.25), re-open the new ones. The README has a gallery of 14 edit types with the prompts and notes on where each LoRA struggles.

Nodes: [https://github.com/PedroMarinhoDev/ComfyUI-Moss-Image-Edit-Native](https://github.com/PedroMarinhoDev/ComfyUI-Moss-Image-Edit-Native)

chinoll LoRA converted for ComfyUI: [https://huggingface.co/PedroMarinhoDev/Moss-Image-Edit-ComfyUI](https://huggingface.co/PedroMarinhoDev/Moss-Image-Edit-ComfyUI)

document fetched 2026-10-25 00:29 · canonical https://reddit.com/r/StableDiffusion/comments/1xem0d0/update_comfyuimossimageeditnative_now_supports