Flux Kontext Max Multi: 6 Two-Image Edits

Multi-image editing gets interesting when a prompt can pull details from a second reference without wrecking the base photo. These examples use Flux Kontext Max Multi to do logo placement, hair color transfer, pattern application, and object-level rearrangements.

Model link

https://wiro.ai/models/wiroai/flux-kontext-max-multi

Test setup

Prompt upsampling: false
Safety tolerance: 2
Output format: jpeg

Edit 1: Put a logo on a shirt (photo + logo)

Input image 1 (base photo).

Input image 2 (logo reference).

Prompt: Put the logo onto the man’s shirt with small size.

This is a clean test for multi-image grounding. A good output keeps folds, lighting, and perspective from the base photo while integrating the logo from the reference.

Edit 2: Hair look transfer (photo + color reference)

Base portrait input — Input image 1 (base portrait).

Hair color reference — Input image 2 (hair reference).

Edited output with hair changed — Prompt: Make the woman’s hair look like it does in the second image.

When this works, the face stays the same and only hair changes. If the output shifts identity, tighten the prompt to say “keep face, pose, and background unchanged”.

Edit 3: Apply a pattern to a wall (room + pattern)

Room photo input — Input image 1 (room).

Pattern reference input — Input image 2 (pattern).

Edited output with patterned wall — Prompt: The first image is the room. The second image is the pattern. Paint the wall behind the bed with the pattern in the second image.

This is one of the best use cases for multi-image edits: keep the geometry and lighting from the room, but borrow surface detail from the second image.

Edit 4: Logo placement on a stylized portrait (illustration + logo)

Stylized portrait input — Input image 1 (stylized portrait).

Input image 2 (logo reference).

Prompt: Print the logo in small size at the center of the woman’s shirt.

Stylized inputs test whether the edit follows the local style. The best outputs keep the illustration look and treat the logo like it belongs there.

Edit 5: Add a sleeve patch (photo + logo)

Base portrait input reused — Input image 1 (base portrait).

Input image 2 (logo reference).

Prompt: Add the logo from the second image as a small patch on the left sleeve of the woman’s hoodie. Keep everything else the same.

This kind of precise placement is where Kontext-style editing stands out compared to generic img2img. It should touch one region and leave everything else intact.

Edit 6: Remove an object + reposition another (two-image guidance)

Tea set scene input — Input image 1 (scene).

Edge guidance reference — Input image 2 (guidance reference).

Edited output with objects removed and moved — Prompt: Remove the tea glass. Put the bird left of the teapot.

Multi-step prompts can work well, but they are sensitive. If one instruction fails, split it into two runs (remove first, then move) and keep the wording short.

Quick takeaways

Use case	Why multi-image helps	Prompt tip
Brand/logo placement	Reference image can anchor shape and style	Say size and exact location
Pattern transfer	Borrow surface detail without changing geometry	Name the target region in plain words
Identity preservation	Base photo can stay stable while details change	Explicitly say what must stay unchanged

Try it

Run Flux Kontext Max Multi on Wiro