Click a thing in your picture and say what to change. Qwen Image 2.1 repaints only that part. The rest of your picture stays exactly as it was.
Watch it used, start to finish: the video on YouTube.
Why use this workflow
Only the part you pick changes. A plain Qwen edit repaints the whole picture, so other things can change with it. Here everything outside the picked part stays your own pixels.
You click, you don't paint a mask. The thing lights up under your pointer, and one click picks it. SAM 3 finds its outline.
Small things get full detail. The picked part is worked on at about 1.5 megapixels and put back into your picture at its own size.
How to use it
Load your image. Optional: drag the blue handles to crop it first.
On Mask by Touch, click the thing you want to change. It lights up before you click.
In What to change, say it the way you would ask a person: "Replace the cup with a glass of orange juice."
Run.
Before and after slides between your picture and the result.
Faster, optional
Download the Viggle Turbo file below and switch Fast on. It runs 8 steps instead of 25, and a run takes about half the time.
What you need
AusBoss nodes 2.11.0 or newer. In ComfyUI-Manager, search "AusBoss". Already have them? Update first. If Manager still shows an older version, pick 2.11.2 in its version list.
ComfyUI 0.38 or newer.
The Qwen Image 2.1 models:
qwen_image_2.1_int8_convrot.safetensors →
models/diffusion_models· 7.26 GBqwen3vl_8b_int8_convrot.safetensors →
models/text_encoders· 9.35 GBqwen_image_2.1_vae_bf16.safetensors →
models/vae· 676 MB
SAM 3, which finds the thing you click: sam3.1_multiplex_fp16.safetensors →
models/checkpoints· 1.75 GB
Optional, for Fast: Qwen-Image-2.1-viggle-turbo-v0.3-6step-lora-r128.safetensors → models/loras · 680 MB. The workflow runs without it as long as Fast stays off.
Tips
Wrong thing picked? The arrows under the picture swap it for the next thing it could mean. Paint adds an area by hand, and Erase takes some away. Before AusBoss nodes 2.12.0 this was one button, Draw.
Pick all of it. A second click adds a piece: the other sleeve, a strap. What you leave out stays.
The new thing is painted inside the pick. Ask for something much bigger and it is cut at the edge. Raise Grow under the picture to give it room.
To remove something, pick it and write "Remove the cup." Pick its shadow too, or the shadow stays.
To add something, switch to Paint (Draw before 2.12.0), paint where it should go with some room, and say how big: "Add small black sunglasses on her eyes."
A hand that was under the old thing can come out odd. Run it again. The seed changes each run.
About 12 to 15 seconds a run on an RTX 5090. A new picture needs about 20 seconds before things light up.
On Krea 2 instead: Krea 2 Inpaint Masked. To take a thing out and keep it as a transparent PNG: Qwen Image 2.1 Cut Out + Remove.
Credits
Qwen Image 2.1 is by Alibaba Qwen, under the Qwen Research License: non-commercial use only. SAM 3 is by Meta, under the SAM License. The optional speed LoRA is Viggle Turbo v0.3 by Viggle, under the Qwen Research License: research use only.
v1.0 (2026-10-13): first release.
Questions or bugs: GitHub issues or the comments here. Videos on YouTube @ausbossdev. More from me on X @Zanzibased and GitHub.
Description
First release. Needs ComfyUI 0.38 or newer and AusBoss nodes 2.11.0 or newer: already have them? Update them in ComfyUI-Manager first, then restart ComfyUI. If Manager still shows an older version, pick 2.11.2 in its version list.
FAQ
Comments (11)
Crazy GOOD!!!!!!!!!!!!!!!!!!!!! IT SHOULDN'T BE FREE! BEST CONTENT ON CIVITAI !
thank you lukas! yep its still free, thank you again for your donation though
genius
The workflow is very convenient, but there are noticeable seams where the masking and inpainting were applied. Which parameters should I adjust? I have uploaded an image showing the issue. (Tone match
0.50
❋
Fix edge halo
on)
i cant see your image,so this is a guess. i’d grow the mask and feather it. on Mask by Touch raise Grow (it starts at 16) so the mask reaches well past the edge of what you picked, then add a little Soften to blend it in. leave Fix edge halo and Tone match alone, set them back to off and 0. let me know if it helps. i also made a video that shows me using this workflow start to finish: https://youtu.be/iMgaiXoB0v4
@AusBoss
Thanks for the reply. I’ve updated the image and re-uploaded it—it should be visible now. It’s also possible that I misunderstood how it works; I was trying to use this workflow to remove clothing, but it seems it’s actually designed for replacing objects....😋
Great,but how do i input multiple images?
this inpaint one here only takes the one picture. i do have a multi image workflow, its more for pulling things from other pictures into your main one. check it out and let me know if this is what you are looking for or not.
https://civitai.com/models/3000002/qwen-image-21-turbo-multi-image-edit
I absolutely love your Qwen Image 2.1 Inpaint workflow! It’s really user-friendly and convenient to work with. ❤️
I was wondering if you could make a character sheet workflow based on it. It would be awesome to upload 1–4 reference images of a character and generate a single sheet showing that same character from different angles, while keeping their face, hairstyle, outfit, body proportions, and overall look consistent.
It would also be great if we could choose how many reference images to use, load our own LoRAs, and have an upscaling option to improve the final image quality and resolution.
Your workflow makes inpainting so convenient, and I think something like this would be incredibly useful for creating consistent, high-quality character sheets. Thanks for sharing your work! ❤️
I have been wanting to make one of these and I have been working on a workflow/lora that takes one picture of a person and will rotate them any direction, so technically you could upload one pic and it would rotate them different angles. I have had mixed results though and have not been super happy with it yet. i probably do have something i can release. its just not 100% accurate. I'm glad you commented this though! this confirms that i need to keep working on that and get something out, i also like your idea of inputting multiple images of them. thats a good plan. I'll try to at least get a version 1.0 up and i can always try to make it better later.
@AusBoss Thats great, all your workflows are perfect and current character sheets make little to no sense. We will wait for your discoveries



