MOSS-Image-Edit-HQ

MOSS-Image-Edit-HQ

An unofficial community LoRA that edits images from written instructions on top of MOSS-Image-SFT, for ComfyUI. For image makers who want to test open instruction-based editing.

Made by akhaliq on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: MOSS-Image-Edit-HQ is an image-editing LoRA built on OpenMOSS-Team/MOSS-Image-SFT, for use in ComfyUI. You give it a photo and a written instruction, and it changes the picture. The author says it is an unofficial community fine-tune with no tie to the OpenMOSS team.

Who it suits: ComfyUI users who want to try instruction-based image editing, and tinkerers who compare open editing models.

Industry | Who uses it | How they use it | Example
Design | Graphic designers | Change colors or clothing in a photo by instruction | Make a jacket red
Marketing | Content teams | Adjust props or objects in product images | Add an object to a scene
Photography | Photo editors | Try quick edits that leave the rest of the scene alone | Change a prop and keep the background
Research | Image-model hobbyists | Compare editing checkpoints side by side | armV5 against the older armV6

You can skip it if you do not use ComfyUI, or if you need an official, supported model.

Where to use it:

• ComfyUI, with native support in a build newer than v0.42.1
• A ComfyUI plugin from Chenkin-x, which uses the LoRA in the other file format
• A demo Space on Hugging Face, with a quality mode and an 8-step fast mode

Use case | Who | What they get back
Instruction edits | Image makers | An edited version of your photo
Color and clothing changes | Designers | The card says these edits worked well
Object additions | Designers | Added objects, which the newer armV6 handled worse

How to set it up:

• ComfyUI newer than v0.42.1, from master or a nightly build
• The base MOSS-Image model file published by Comfy-Org
• The armV5 checkpoint, which the card recommends

1. Update ComfyUI past v0.42.1.
2. Put the LoRA file into the ComfyUI models loras folder.
3. Add a LoRA loader after Load Diffusion Model.
4. Add the Text Encode MOSS-Image Edit node, connect the VAE and reference images, and set edit_format to ai-toolkit.

pip install -U huggingface_hub
hf download akhaliq/MOSS-Image-Edit-HQ

How to use it: Open Browse Templates and pick the MOSS-Image Edit template, then drop the LoRA into its loader. Set edit_format to ai-toolkit, add your image and type an instruction. The card uses 50 steps, CFG 3, Euler sampler and SLG layer 22.

edit_format = ai-toolkit

Example: You load a portrait and write an instruction to change the shirt color. The workflow returns the edited photo with the rest of the picture preserved, as the card's showcase examples do.

For newcomers and experts: If you are new, use the built-in template and the armV5 file, and test on a photo you do not mind altering. References are downscaled to 384 by 384 automatically.

If you know the field, armV5 was trained with full scene captions, a masked loss with an inverted mask prior, and preservation contracts. The author reports that armV6 did not improve on it and introduced subtle color casts, and keeps it for transparency.

Load Diffusion Model
LoRA loader with armV5
Text Encode MOSS-Image Edit
KSampler
VAE Decode and Save

Good to know:

• License: Apache 2.0
• Cost to run: your own GPU hardware, with no paid service named
• Data it reaches: the images you edit and the files you download
• Not for: official use, since it is unaffiliated with the OpenMOSS team
• Limits: the card says older ComfyUI builds can hit errors, with v0.42.3 as the tested baseline

On Hugging Face, 13 November 2026

• Library: comfyui
• Downloads in the last 30 days: 0
• Likes: 90
• License: apache-2.0
• Based on: OpenMOSS-Team/MOSS-Image-SFT
• Tags: comfyui, lora, image-editing, moss-image, ml-intern

Links

• Model: https://huggingface.co/akhaliq/MOSS-Image-Edit-HQ
• Files: https://huggingface.co/akhaliq/MOSS-Image-Edit-HQ/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download akhaliq/MOSS-Image-Edit-HQ

Then follow the usage in its model card. Nothing in this kit runs by itself.