Combining Multiple Images with Nano Banana in Kaiber Canvas

Got images you want to merge into something new? Nano Banana models in Canvas (Nano Banana, Nano Banana Pro and Nano Banana 2) support uploads of up to 10 images to use as references to create a new image. 

How to combine images in Canvas

Right-click anywhere on an open canvas to bring up the menu and select Create Image, or from the Canvas home page select Text to Image from the image menu. Click the model name and select Nano Banana, Nano Banana 2, or Nano Banana Pro.

For combining images, Nano Banana 2 and Nano Banana Pro are the stronger options. They handle more complex prompts and can manage multiple changes across several references in one generation.

Upload your images into the image area, then write a prompt that references them by position: e.g. the person from image 1 wearing the clothing from image 2, in the setting from the background from image 3. The more specific you are, the better the result.

Adjust the aspect ratio before generating to fit the format you’re working toward.

What you can do

Merge subjects and backgrounds. Drop a subject into a completely different scene, or blend two environments together.

Add a logo or graphic to a product. Place a logo onto packaging, apparel, or a product shot without touching your original files.

Swap clothing or props. Reference an item from one image and place it onto a subject from another.

Build album art from references. Combine an artist photo, a location, a mood reference, and typographic style into a single generation.

Create group compositions. Pull multiple people from separate photos and place them together in a shared scene.

Layer style and mood. Feed in several aesthetic references at once to define a colour palette, lighting style, and composition in one prompt.

Build storyboard sequences. Use edited frames as references to maintain visual continuity across a series of images — then animate them into a video using models like MiniMax, Veo, or Wan.

A few tips

Be specific about placement and relationships between images. Vague prompts get vague results. If something looks off, tweak the wording and generate again. Small changes to phrasing can make a real difference.

Clean source images work best. A subject on a simple background, a logo on white, a product shot with good lighting. The more the model has to wrestle with a messy reference, the harder it is to get a clean output.