FLUX 3 Image to Image: Edit, Restyle and Combine Photos

Last verified: November 1, 2026

FLUX 3 reworks the photos you upload based on a written instruction. It can return the edit in a 7:5 landscape or 5:7 portrait frame, which is the shape of a standard 5×7 print in either direction. That makes FLUX 3 image-to-image editing a quick way to turn a phone snapshot or product shot into a picture already shaped for a frame, a greeting card or a printed portfolio. Square, widescreen and vertical formats are there too when a job needs them.

FLUX 3 can edit images in a wide range of styles, aspect ratios and resolutions. When you edit, each upload becomes a numbered reference: image 1, image 2 and so on. You tell the model in plain language what each one adds to the result, so a single prompt can draw on several pictures at once.

FLUX 3 takes everything you want to control from the prompt, so the Prompt box does most of the work here. The other settings set the frame and the output size, and decide whether the model looks up real-world references online first. A clear instruction plus two or three deliberate choices is the whole job.

Explore more image to image

Image To Image About Flux 3 Use in Text To Video Use in Image To Video Use in Reference To Video Use in First To Last Frame Use in Extend Video Seededit V3 Seedream 4.5 Edit Nano Banana 2 GPT Image 2 Qwen Image 3.0 Pro Qwen Image 3.0
Capability Snapshot

FLUX 3 Editing Specs at a Glance

What goes in, what you can set and what you get back from each FLUX 3 edit.

Input

Up to 10 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, 3:2, 2:3, 16:9, 9:16, 21:9, 9:21, 5:4, 4:5, 7:5, 5:7, 2:1, 1:2

Output

1 image per run

How Do You Get Better Edits From FLUX 3?

Reframe a Shot Into a New Shape

Sometimes the edit should change the frame, like turning a square product photo into a wide banner or a print-shaped portrait. In that case, pick the target ratio in Aspect Ratio instead of leaving it on its default. Then say in the prompt what should fill the new space, such as more of the marble counter, open sky or the rest of the shelf. The default keeps the source's frame. That's right for a straight retouch, but a reframe works better when you name the extra content, so you aren't leaving the model to guess it.

Stack Several Changes Into One Edit Record

For an edit with several parts, write the prompt as a record. Name the image and the operation, give each change with its location, then list what stays. For example: in image 1, change the mug on the left to matte black, replace the poster on the back wall with a plain linen panel, keep the person, the lighting and the window view unchanged. A structured instruction like this also works as a checklist when you compare the result with your source.

Pin Exact Colors and Wording

When a color has to match a brand swatch, give it as a hex code like #1F5C5B instead of a name like dark teal. Put any new words the image should show in quotation marks, word for word. These two habits matter most for packaging, signage and poster edits, where a near-miss color or a reworded slogan means doing the job again.

Editing an Image With FLUX 3, Step by Step

Five steps take you from upload to a downloaded edit.

1

Upload your source in "Add Image"

Start with the picture you want to edit, then add more images only if the instruction uses them, such as a garment, a product or an object to place in the scene. Order matters. FLUX 3 calls the first upload image 1, the next image 2 and so on, so put the photo you are editing first and refer to the others by number in the prompt.

2

Write the edit in "Prompt"

Write the change as an instruction and name exactly what should change. For example, 'make the mug matte black' works better than 'improve the mug'. If you uploaded several images, say what each one adds. The box takes up to 5,000 characters, far more than most edits need. If you only have a rough idea, "Enhance Prompt" under the box suggests fuller versions, and your text stays as written until you pick one.

3

Choose "Aspect Ratio" and "Resolution Quality"

"Aspect Ratio" starts on "Auto", which keeps your source's frame and is usually right for a retouch. The fixed choices are 1:1, 4:3, 3:4, 3:2, 2:3, 16:9, 9:16, 21:9, 9:21, 5:4, 4:5, 7:5, 5:7, 2:1 and 1:2. Pick 7:5 or 5:7 when the edit is headed for a landscape or portrait print. "Resolution Quality" offers "1K" (the default), "2K" and "4K". Get the instruction right at 1K, then switch to 4K for the version you keep. Check that 4K render on its own, because it is a new generation.

4

Start the run with "Generate"

Before you click "Generate", check "Web Search". It starts switched on and lets the model look up references online before it works on your edit. The FAQ below explains when to turn it off. Each run returns a single image, so expect a few runs when you are testing different wordings.

5

Download the result

Download the image once it's ready and compare it side by side with your source before deciding what's next: a tighter instruction, a different frame, or a follow-up edit on the new file.

Choosing Between Editing an Upload and Starting From Text

Two ways to use FLUX 3 for the same picture, depending on what you already have.

Starting point Editing starts from the photos you upload and changes them according to your instruction, so anything already in the frame can stay. The same model's Text To Image workflow builds the whole scene from your prompt, which suits ideas that don't exist as a photo yet. Choose image-to-image when you already have the shot, product or person that has to stay, and text-to-image when the picture has to be invented from words.
How you write the prompt The prompt is an instruction that names exactly what to change, and with several uploads it says what each one adds. In text-to-image the prompt describes the subject, setting, light and framing, because the model creates the entire scene from it. If you think in terms of fixes to an existing picture, stay with editing; if you'd rather describe a scene from scratch, text-to-image fits better.
Use FLUX 3 for editing when an existing image needs to carry through the change. Start from text when there is nothing to edit yet.

Why Is My FLUX 3 Edit Not Coming Out Right?

Common problems with image edits, and what to change before the next run.

The edit changed parts of the picture I wanted to keep

If an instruction never says what must stay, those parts are left open to change. Rewrite it the way Black Forest Labs' own edit examples do: name the target by what it is ('the large tiger', not 'the animal'), state the change once, then list what must not change, such as composition, lighting and background. Run it again from the same source image.

I wrote 'no crowds' but people still show up

A negative instruction can produce an unwanted result, sometimes the very thing it names. Describe the scene you want directly instead, like 'an empty promenade' rather than 'no crowds', and generate again with the rewritten prompt.

The result came out in the shape of the wrong reference photo

With Aspect Ratio on its default, the output takes the shape of the first image you uploaded, so a portrait reference placed first gives a portrait result. Re-add your images with the one whose frame you want at the front, or pick a fixed ratio, and run the edit again.

My image won't upload

Add Image only takes files that meet its format, file-size and minimum-dimension limits, which are listed in the upload FAQ below. Re-export the picture in one of the listed formats, compress it if it's too heavy, or use a larger original if it's too small, then add it again. Nothing needs to change in your prompt or settings.

My detailed brief doesn't fit in the prompt box

The Prompt box has a character limit (the number is in the steps above). Very long briefs often re-describe things your uploads already show. Cut the text down to the change itself, what each upload adds and what must stay the same, then generate again.

Frequently Asked Questions

Can I keep editing the same image over several rounds with FLUX 3?

Yes. Black Forest Labs presents FLUX 3 as capable of precise multi-turn edits that leave every other pixel unchanged, which suits a step-by-step retouch. To work this way, download each result, add it back through Add Image as the first image, and write only the next change in the prompt. Making one change per round also makes it easy to see which instruction caused a problem.

How many images can I upload, and which files work?

Add Image needs at least one image and takes up to ten per run. It accepts JPG, JPEG, PNG and WEBP files of up to 10 MB each, and every image must be at least 300×300 pixels. Put the photo you're editing first and add the rest only when your instruction uses them.

Why does one generation cost more than another?

The credit cost of a run is calculated from the options you choose and appears above the Generate button once you're signed in. Resolution Quality is the setting that changes it, so check the figure again after you switch output sizes.

What does the Web Search switch do with FLUX 3?

It lets FLUX 3 search the web before generating so that real things come out right, which helps when an edit involves a real landmark, product or event. Black Forest Labs shows this with a request for an official festival poster: with search off, the model invented a design, and with it on, it produced the actual poster. Turn it off when your uploads and instruction already contain everything the edit needs. Results then come faster and are based only on your prompt.

References

Sources and citations used to support the content provided above.

Updated: 2026-11-01 13:37:05 3 Sources

docs.bfl.ai

Source Link
https://docs.bfl.ai/flux_3/flux3_image_layout

docs.bfl.ml

Source Link
https://docs.bfl.ml/guides/prompting_unified_basics

docs.bfl.ai

Source Link
https://docs.bfl.ai/api-reference/utility/generate-an-image-with-flux-3