FLUX Multi Image – Combine Multiple Images into One Picture

FLUX.1 Kontext was built around one input image, so multi-image work with Kontext usually meant stitching pictures side by side or using experimental multi-image endpoints. FLUX 3 Image handles multiple images directly: you upload as many as ten references, refer to each by number, and say what each one contributes to the result.

That is why this page runs FLUX 3 Image rather than Kontext. If you searched for "FLUX Kontext multi image" or "FLUX Kontext multiple images", you most likely want exactly this: several pictures in, one coherent picture out. Kontext is still the right tool for quick single-photo edits, and has its own page.

These are the three references for the next two examples. All three were generated as plain product shots on white: a brass lamp (image 1), a rust bouclé armchair (image 2) and a round geometric rug (image 3). Isolated products like these are the easiest references for FLUX 3 to read, because nothing in the background competes with the object.

The prompt named each image by number: the armchair from image 2 by a tall window, the lamp from image 1 on a small oak side table, the rug from image 3 under the chair, afternoon sun. FLUX 3 Image kept the pleated shade, the brass stem, the walnut legs and the blue diamond pattern of the rug, and lit all three with the same window light.

Notice what it added on its own: the side table, a mug, a plant on the windowsill, a floor vent. Anything you don't specify, FLUX 3 fills in to make the room believable. If you want fewer extras, say so ("no other furniture").

Same three references, new prompt: a dark green study at night, the lamp switched on as the only light source, the armchair angled toward a bookshelf, the rug underneath. We also switched the aspect ratio to 4:5 for a social feed.

The products still match their references, while the light, wall color and camera angle are new. This is how one set of product photos turns into a whole campaign: keep the references, change the scene description.

1. Upload the references in a sensible order. Image 1 sets the output shape when aspect ratio is on auto, so put the main scene or hero product first.
2. Refer to each image by number. "The armchair from image 2" is clearer to FLUX 3 than "the chair".
3. Say what each image contributes. An object, a person, a background, a color palette or a style.
4. Place things in the scene. Left, right, on the table, under the chair, in the background.
5. Describe the light once, for the whole scene, so every object looks like it belongs in the same room.

• Product plus setting. A packshot and a photo of a kitchen, a beach or a shelf.
• Several products. A full outfit from separate item photos, or a furniture set from separate listings.
• Character plus props. One character reference plus objects they should hold or wear. See FLUX consistent character.
• Content plus style. A photo plus a painting or print whose look you want. See FLUX style transfer.
• Moodboard to image. Four or five images that share a palette or atmosphere, plus a short scene description.

• Use clear references. One object per image, well lit, not cropped at the edges.
• Keep it to what you need. Every reference adds credits, and FLUX 3 balances ten inputs less precisely than three. Start with the fewest images that do the job.
• Fix one thing per re-run. If the lamp came out the wrong size, describe its size and leave the rest of the prompt alone.
• Use the result as a new reference. A good result from one request can be image 1 of the next.