Compose multiple images

The Compose multiple images mode (Compose) merges 2 to 6 images into a single coherent scene.

This is the right workflow when you want a “collage, but natural” result: subjects and objects from different photos are integrated into one composition with consistent light, scale, perspective, and style.

In this mode the prompt is still the best way to control the final scene, but it is no longer mandatory: if you leave it empty, the app uses internal composition guidance to keep the main subjects and integrate them naturally.

Use Compose multiple images if:

• you want to combine multiple people/objects/products into one image;
• you want to build a scene from different shots (even with different angles);
• you want to compose a marketing hero shot with multiple elements;
• you want to experiment with multiple subjects without manual cutouts.

• If you need manual control over cutouts and placement (scale/rotation/position): use Seamless Blend (subjects).
• If you want to edit a single photo without compositing: use Single image edit.
• If you want to generate an image with no inputs: use Generate image from text.

• Minimum 2 images, maximum 6 (app limit).
• Formats: PNG/JPG/GIF (as per upload support).
• Optional prompt: recommended when you want stronger control over layout, atmosphere, and subject priority.

When you start processing, the app sends to the model:

• the images you uploaded (up to 6);
• your prompt, if present, or a default composition guide;
• any available settings (e.g. ratio/resolution if supported by the model).

The model returns a single “final” image, trying to:

• harmonize light and colors between different inputs;
• make perspective consistent;
• avoid cutout edges and obvious artifacts;
• build (if needed) a background consistent with the scene.

In this mode you typically see:

• Mode: workflow selector (choose Compose multiple images).
• Multi upload: upload 2-6 images.
• Prompt: optional, but useful to describe the final scene, scale, and atmosphere.
• Compose: start processing.
• Generate variants (2-4): create alternatives (very useful: quality varies a lot between variants).
• Use as input: for later refinements with other workflows (often Edit or Seamless Blend).
• Once you reach 6 images, upload is blocked until you remove one.

1. Choose compatible images (same light type and angle when possible).
2. Write a prompt that defines:
   • main subjects,
   • layout (left/right, foreground/background),
   • relative scale,
   • atmosphere and style.
3. Generate the output.
4. Generate 2-4 variants and pick the best.
5. Refine:
   • Single image edit for micro corrections,
   • or Seamless Blend if you need to fix placement/cutouts.

In the “Mode” selector, choose Compose multiple images.

Practical tips for choosing inputs:

• If you combine people, prefer photos with similar lighting (all indoor or all outdoor).
• If perspective is very different (wide vs tele), integration may look less natural.
• If one image has a complex background and another is studio, specify in the prompt which style you want for the final scene.

The prompt should tell the model:

• what must be present,
• where it should be,
• how big it is relative to others,
• what scene should result.

Simple example: “Create a realistic scene: person A on the left, person B on the right, both standing, same level of soft studio light, neutral gray background, professional photographic look.”

Click Compose and evaluate:

• scale coherence (one should not be huge compared to the other without reason);
• clean edges (no halos or cut borders);
• consistent light/shadows;
• overall quality.

Use Generate variants (2-4) and pick the most natural:

• often one variant handles edges and light better;
• another may have a nicer composition.

The best refinements are almost always done in two ways:

• Single image edit: micro corrections (artifacts, details, cleanup).
• Seamless Blend: if you want to reposition a subject or fix a cutout in a controlled way.

“Merge the images into one realistic photo. Composition: [LAYOUT]. Scale: [RELATIVE SCALE]. Lighting: [LIGHT]. Background: [BACKGROUND]. Keep a natural look, clean edges, no artifacts.”

“Create a clean product image: arrange the products neatly on a studio background, softbox light, realistic contact shadows. Accurate colors, no text or watermark.”

“Create a realistic lifestyle scene: [SCENE]. Subjects should look in the same place, with consistent light and harmonized colors.”

• “same light level / same light direction”
• “coherent shadows”
• “realistic scale”
• “clean edges, no halos”
• “realistic photographic look”

• Even without a prompt, Compose tries to keep the main subjects from each input, align scale/perspective/light, and avoid duplicating subjects.
• If the prompt is vague or missing, the output can still work well for simple merges but be less precise about layout and priority.
• If you want geometric control (precise position/rotation), Compose is not the right tool: use Seamless Blend.
• If one image contains text/complex logos, they can degrade: consider a refinement in Edit.

• Specify scale: “person A and B same height; product X 20% smaller than product Y”.
• If you need precise control, switch to Seamless Blend.

• Generate variants: often one resolves it.
• Then refine in Single image edit: markup on halos + “clean up edges”.

• Write explicitly: “light direction from the left, soft; light and coherent shadows”.
• Choose more compatible inputs (same light).

• Specify the background: “neutral studio”, “white wall”, “urban outdoor”.
• If you want a precise context, describe it like you would instruct a photographer.