Change a picture with one or two more pictures to take from. It is made for clothes: dress the person in your picture in what someone wears in another picture, with the official Qwen Image 2.1 Turbo checkpoint, in 8 steps.

How to use it

1. Load the picture you want to change as Image 1.

2. Load the picture to take from as Image 2.

3. Write what to take: "Put the outfit from image 2 on her."

4. Run.

It opens set up for clothes: Outfit swap and Auto prompt are on, so a few words are enough. For another kind of edit, switch Outfit swap off and write what to do: "Give the woman in image 1 the haircut of the woman in image 2."

The four switches

• Image 3. One more picture to take from. Call it image 3 in the prompt: "the jacket from image 2 and the skirt from image 3".

• Auto prompt. Qwen looks at your pictures and writes the full instruction from a few words of yours. What it wrote shows in Show Text. It uses the text encoder the workflow already loads, so there is nothing more to download.

• Outfit swap. For changing clothes. My Outfit Swap LoRA takes over from my Consistency LoRA. With Auto prompt on too, Qwen names each piece of clothing and puts it into the sentence that LoRA knows.

• Upscale the result. Redraws the result bigger in tiles with my Restore LoRA, to the Megapixels you set.

Auto prompt and Outfit swap are on as the file opens. Image 3 and Upscale the result are off.

What you need

• AusBoss nodes 2.10.0 or newer. In ComfyUI-Manager, search "AusBoss". Already have them? Update first.

• ComfyUI 0.38 or newer.

• The models:

  • qwen_image_2.1_turbo_int8_convrot.safetensors → models/diffusion_models · 7.26 GB

  • qwen3vl_8b_int8_convrot.safetensors → models/text_encoders · 9.35 GB

  • qwen_image_2.1_vae_bf16.safetensors → models/vae · 676 MB

• My Outfit Swap LoRA, for Outfit swap: qwen-image-2.1-outfit-swap.safetensors → models/loras · 159 MB

• My Consistency LoRA, for edits with Outfit swap off: qwen-image-2.1-consistency.safetensors → models/loras · 159 MB

• Only for Upscale the result, my Restore LoRA: qwen-image-2.1-restore.safetensors → models/loras · 159 MB

Tips

• Name the clothes, or let Auto prompt name them. "The outfit from image 2" alone changed nothing in 7 of my 18 test swaps. With each piece named, all 18 went through.

• A wrong word gets drawn. If a piece comes out with the wrong color or pattern, read Show Text. Copy the sentence into your prompt, fix the word and run with Auto prompt off.

• Write a place as a change to Image 1: "Change the background of image 1 to the street from image 2." Asking to put her into image 2 gives back image 2.

• Crop a picture to the thing you want taken from it. Drag the blue handles.

• One picture can hold several things. A jacket and a skirt side by side work as one picture.

• Image 1 runs at 2 megapixels and the others at 1. Swaps came out best that way. To change it, set Megapixels on a picture node. For a bigger result, use the Upscale switch.

• Shoes only go on when Image 1 shows the feet. With Outfit swap and Auto prompt on, the workflow checks that itself.

• Hair and light work too, with Outfit swap off.

• Times on an RTX 5090. About 15 seconds as it opens with two pictures, about 20 with three. The upscale to 8 megapixels adds about 30.

What it does not do

• A pose from another picture does not carry over. The person keeps the pose of Image 1.

• It is not for swapping people. For a face or a whole person from another picture, use my Qwen Image 2.1 Person Swap.

• The upscale is a redraw. Fine detail is made up to fit, and small lettering can come back misspelled.

• Everything was measured on one RTX 5090.

For one outfit picture on the regular 25-step model, see my Qwen Image 2.1 Outfit Swap. For new pictures and one-picture edits with the Turbo checkpoint, see Qwen Image 2.1 Turbo Generate + Edit.

Credits

Qwen Image 2.1 Turbo is by Alibaba Qwen, under the Qwen Research License: non-commercial use only. The ComfyUI files are from Comfy-Org.

v1.0 (2026-10-13): first release.

Questions or bugs: GitHub issues or the comments here. More from me on X @Zanzibased and GitHub.