Flux 2 Max Image to Image: Multi-Reference Photo Editing

Last verified: October 29, 2026

Flux 2 Max changes the photos you upload based on what you write in the prompt. Along with the standard ratios, it can return the edit in an ultrawide 21:9 frame or an ultra-tall 9:21 frame, so you can build the result for a website banner or a full-screen story from the start. Flux 2 Max image-to-image editing can start from one photo or from a set of references, and your written instructions control every change.

Black Forest Labs built Flux 2 Max for the highest editing consistency across tasks and the strongest prompt following of its FLUX.2 models. Both qualities matter for photo edits. Most jobs mean changing one thing, such as a background, a surface or a line of text, while the rest still looks like the same shot. Your instruction also has to be followed as written, not loosely interpreted.

The workflow is short: add your photos, write the instruction, choose a frame and a resolution, then download the finished image. The sections below cover which settings to start with, how to word an edit so the model reads it correctly, and what to do when a result misses.

Explore more image to image

Capability Snapshot

Flux 2 Max Editing Specs at a Glance

The inputs, frame options and output you work with in this editor.

Input

Up to 8 images (JPG, JPEG, PNG or WEBP, up to 10 MB each) + a text prompt of up to 5,000 characters

Resolution

1K, 2K or 4K

Aspect Ratio

Auto, 1:1, 4:3, 3:4, 16:9, 9:16, 3:2, 2:3, 21:9, 9:21

Output

1 image per run

Prompting Tips for Sharper Flux 2 Max Edits

Lead the Prompt With the Change That Matters Most

Flux 2 Max pays most attention to what comes first in a prompt. Open with the element you are changing and the result you want, then add style and context. The recommended order is main subject, key action, critical style, essential context, then secondary details. For an edit, write 'Replace the sofa with a rust-orange velvet sofa, morning light from the left'. Don't bury the sofa behind a string of adjectives.

Quote the Exact Words When Editing Signs and Labels

Flux 2 Max can edit text inside a photo. Lettering comes out most reliably when the exact words are in quotation marks and you describe where they sit and how they look, for example: 'Change the shop sign to read 'NORTHLIGHT' in bold industrial lettering, centered above the door.' Give the text a size role too, such as large headline or small body copy, so the new wording fits the scale of the sign or package it replaces.

Pin Exact Colors With Hex Codes

When a recolor has to match a brand swatch, put the hex code after the word 'color' or 'hex' and tie it to one named object, as in 'the sofa in color #1B6B6F'. Hex codes work best when each one is attached to a specific object. A loose request like 'use #FF0000 somewhere' can give inconsistent results. With several brand colors, give each one its own object so none of them drifts.

How to Edit an Image With Flux 2 Max

Five steps take you from upload to download.

1

Add your photos in "Add Image"

Start with the photo you want to change. Then add any references the edit should draw from, such as a fabric pattern, a garment or a second product. The form needs at least one image before it can run. Leave out pictures that have no job in the edit: every upload should have a role you can name in the prompt.

2

Describe the edit in the "Prompt" box

Write what should change and what the finished image should look like. The box takes up to 5,000 characters, which is far more than an edit usually needs. If you only have a rough idea, click "Enhance Prompt" under the box to see fuller versions. Your own text stays as it is until you pick one of the suggestions.

3

Choose "Aspect Ratio" and "Resolution Quality"

"Aspect Ratio" starts on Auto and also offers 1:1, 4:3, 3:4, 16:9, 9:16, 3:2, 2:3, 21:9 and 9:21. Pick a fixed ratio when the image has a set destination, such as 21:9 for a site header or 9:21 for a full-screen phone story. "Resolution Quality" defaults to 1K, with 2K and 4K above it. Test the prompt at 1K and switch up only for the version you plan to keep, so the large render happens once the edit is right.

4

Click "Generate"

Each run returns one image. If it misses, change one thing at a time, like the wording of the change or the role you gave a reference, so you can tell which adjustment fixed it.

5

Download the result

Save the image and check it at full size before you use it. Look closely at faces, the edges of retextured surfaces and any lettering, because flaws there are easy to miss in a thumbnail.

Flux 2 Max or Another FLUX.2 Variant for Photo Edits

How Flux 2 Max compares with the other FLUX.2 models Black Forest Labs offers for editing.

Final assets versus high-volume production Black Forest Labs lists it as best for the highest quality and final assets. FLUX.2 [pro] is the variant Black Forest Labs lists as best for production at scale. Choose Flux 2 Max for hero images and client deliverables that have to hold up at full size, and FLUX.2 [pro] when you are running large batches of routine edits.
Edits where lettering is the main job It has the reliable text rendering that Black Forest Labs documents for every FLUX.2 model, and it is the variant they point to for the highest quality. FLUX.2 [flex] is the variant Black Forest Labs points to for typography and fine control. Keep Flux 2 Max when text is one part of a bigger edit, and look at FLUX.2 [flex] when the image is mostly type, such as a poster or a label layout.
Number of references in one edit It is documented for up to 8 reference images through the API and up to 10 in Black Forest Labs' playground. FLUX.2 [klein] tops out at 4 reference images, both in the API and in the playground. When an outfit, room or product composite needs more than four source photos, Flux 2 Max is the better fit; FLUX.2 [klein] only works for jobs that need four references or fewer.
Use Flux 2 Max when the edited image is the final asset or the composite needs many references. FLUX.2 [pro] suits bulk production, FLUX.2 [flex] suits type-led layouts, and FLUX.2 [klein] suits jobs that need four references or fewer.

Fixing Common Flux 2 Max Editing Problems

What usually goes wrong with photo edits, and what to change before your next run.

I wrote 'no people' or 'no blur' and the edit ignored it

Flux 2 Max does not support negative prompts, so a 'no' instruction gives the model nothing to aim for. Rewrite the line as the state you want. Say 'sharp focus throughout' instead of 'no blur', or 'an empty street' instead of 'no people'. Then run the edit again with the same images.

It took the style or background from the wrong photo

When you upload several images, the model needs to know what to take from each one. Give every image a role in the prompt, for example the subject from image 1, the pattern from image 2 and the background from image 3. Add a few words on what each image shows so the instruction doesn't depend on the number alone. Then generate again.

The edit changed parts of the photo I wanted to keep

A loose instruction such as 'make it look better' leaves it open how much of the photo to rework. Be specific about what changes and clear about the end state. Name the one element to alter, describe how it should look afterwards, and say what has to stay as it is, such as the face, the pose or the product's shape. Then run it again. If several things need to change, test them in separate runs first so you can see which instruction causes the drift.

My photo won't upload

Add Image only accepts certain file formats, a maximum file size and a minimum image size, and it stops taking files once the upload limit is reached. All of these are listed in the upload FAQ below. Convert other formats to one of the accepted ones, compress or resize heavy files, replace small thumbnails with the original photo, and remove a reference you don't need before adding another. Then upload again.

Frequently Asked Questions

Which files can I upload to Flux 2 Max, and how many at once?

Add Image accepts JPG, JPEG, PNG and WEBP files of up to 10 MB each, at least 300×300 pixels, and you can add between one and eight images per run. A prompt is also required, so the form needs at least one image plus written instructions before it can generate.

Why does one generation cost more than another?

The credit cost of a run is calculated from the options you choose. Once you are signed in, it appears above the Generate button, so you see it before each run. Two things change it: the Resolution Quality you pick and the number of images you upload in Add Image. The Enhance Prompt button does not use credits.

Can Flux 2 Max turn a photo into a different art style?

Yes. Describe the look you want in the prompt, such as a gouache travel-poster illustration, a 35mm film still or a soft watercolor. Faithful representation of many different styles is one of the strengths documented for Flux 2 Max. Its character consistency also holds across styles, so a restyled portrait still looks like the same person.

To copy a look from a picture instead of describing it, upload that picture too and say in the prompt that it provides the style, for example 'subject from image 1, style from image 2'.

How long should a Flux 2 Max editing prompt be?

Medium-length prompts of about 30 to 80 words usually work best. Use 10 to 30 words for quick concepts and style exploration, and go past 80 words only for complex scenes that need detailed specifications. The Prompt box holds far more than that, so length is rarely the problem; clarity is. Spend the words on the change and the end result, not on stacked quality adjectives.

Can I write a structured JSON prompt instead of plain sentences?

Yes. FLUX.2 [pro] and [max] support JSON structured prompting, which is recommended for complex scenes and automated workflows. It uses keys such as scene, subjects, style and lighting, and you can include hex colors for each subject. Paste the JSON into the Prompt box as written. It helps most when an edit involves several objects that each need their own description.

References

Sources and citations used to support the content provided above.

Updated: 2026-10-29 21:18:08 6 Sources

docs.bfl.ai

Source Link
https://docs.bfl.ai/flux_2

bfl.ai

Source Link
https://bfl.ai/models/flux-2-max

docs.bfl.ai

Source Link
https://docs.bfl.ai/guides/prompting_guide_flux2

docs.bfl.ai

Source Link
https://docs.bfl.ai/flux_2/flux2_overview

help.bfl.ai

Source Link
https://help.bfl.ai/articles/7655484417-what-flux-models-are-available

help.bfl.ai

Source Link
https://help.bfl.ai/articles/6944273991-what-are-the-flux-2-technical-specifications