Edit one picture, or up to six together, with FLUX 3 Image Klein. Image 1 is the picture that changes. Images 2 to 6 are pictures to take from: clothes, a hat, a place, a second person. With Image 1 alone, your words do the edit.

How to use it

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

2. To take from other pictures, load them as Image 2 and up, and switch each one on in the card. For an edit with words alone, leave them off.

3. In the box over each picture, type what to take from it: "jacket", "straw hat", "outfit".

4. Run.

It opens set up for clothes: Outfit swap and Auto prompt are on. The prompt box is optional: add anything else there, like "keep her own shoes". For another kind of edit, switch Outfit swap off and write what to do in the prompt box: "make it snow", or "Change the background of image 1 to the staircase in image 2."

The switches

• Image 2 to Image 6. One switch for each picture to take from. Switch them on in order: a picture that is off is passed over and the next one takes its number.

• Auto prompt. Qwen3-VL looks at your pictures and writes the full instruction from the boxes or a few words of yours. What it wrote shows in Show Text. Off, your prompt goes to FLUX 3 as typed and the boxes are not used.

• Outfit swap. For changing clothes. With Auto prompt on too, Qwen3-VL names each piece and puts it on the right image number.

Image 2, Auto prompt and Outfit swap are on as the file opens. Image 3 to Image 6 are off.

What you need

• ComfyUI 0.41.0 or newer. That is the first version with FLUX 3.

• AusBoss nodes 2.13.0 or newer. In ComfyUI-Manager, search "AusBoss". Already have them? Update first.

• The FLUX 3 Image Klein files. They are on Hugging Face behind a license click: open black-forest-labs/flux-3-image-klein-comfy, sign in and accept, then these links work.

  • flux3_image_klein_sd_int8_convrot.safetensors → models/diffusion_models · 5.83 GB

  • qwen3vl_4b_int8_convrot.safetensors → models/text_encoders · 4.86 GB

  • flux2-vae.safetensors → models/vae · 336 MB

• For Auto prompt: qwen3vl_8b_int8_convrot.safetensors → models/text_encoders · 9.35 GB. It is Qwen Image 2.1's text encoder, so you may have it already. FLUX 3's own text encoder mixed up the picture numbers in 4 of my 5 tests; this one got all 5 right.

These are the INT8 files, the same ones ComfyUI's own FLUX 3 template uses. The full-size BF16 files are on the same Hugging Face page for bigger cards (11.1 GB and 8.88 GB): pick them in Diffusion model and Text encoder.

Tips

• Call the pictures by number: "the jacket from image 2". When I named the pieces without their numbers, her socks were changed too in 4 of 4 runs; with the numbers they stayed in 8 of 8. The boxes and Auto prompt do the numbering for you.

• Say what stays: "Keep her face, hair, pose and the background exactly the same."

• Write a place as a change to Image 1: "Change the background of image 1 to the street from image 2."

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

• 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.

• Shoes need the feet in Image 1. Leave shoes out when Image 1 does not show the feet.

• Image 1 runs at 2 megapixels and the others at 0.5. The result has the size and shape of Image 1. To change it, set Megapixels on a picture node.

• Times on an RTX 5090: 12 to 15 seconds for most runs, up to about 24 with Auto prompt.

What it does not do

• No upscale. The result comes out at Image 1's size, about 2 megapixels.

• No LoRAs of mine. LoRA Loader is empty, for FLUX 3 Klein LoRAs of your own.

The same on Qwen Image 2.1: my Qwen Image 2.1 Turbo Multi-Image Edit. To extend a picture with FLUX 3: FLUX 3 Image Klein Outpaint.

Credits

FLUX 3 Image Klein is by Black Forest Labs, under the FLUX Kommunity License: use of the pictures you make is covered by its rules for outputs (section 2), so read it if you plan to use them commercially. The FLUX 3 files are Black Forest Labs' own ComfyUI versions. Qwen3-VL is by Alibaba Qwen; its ComfyUI file is from Comfy-Org.

v1.0 (2026-10-15): first release.

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