Change an image without a mask
Upload an image and describe the edit in plain language. Nano Banana 2 can transform the supplied image or use it as a reference.
Nano Banana 2 is Google's fast image model for creating and editing images with strong world knowledge, readable text, up to 14 references, and output choices from 1K to 4K.
Upload an image and describe the edit in plain language. Nano Banana 2 can transform the supplied image or use it as a reference.
Supports up to 14 JPEG, PNG, or WebP input images for reference-led generation and editing.
High-fidelity text rendering and localization support for marketing visuals, posters, and UI assets.
Google's model is designed to use broad world knowledge when interpreting real-world subjects, locations, and visual references.
A required text prompt plus optional image references. Accepts up to 14 JPEG, PNG, or WebP images, each up to 30MB.
1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 8:1, 1:8, 16:9, 9:16, or auto.
1K, 2K, or 4K. Higher resolutions take longer to generate.
JPG or PNG, with JPG as the default.
One generated image per task; the studio issues separate tasks when multiple variants are requested.
Use natural-language transformations, reference images, semantic understanding, and high-fidelity text rendering in one image workflow.
Generate several visual directions quickly, then move the strongest route to a higher-resolution output.
Combine product, style, and composition references without building a separate masking workflow.
Use the documented text rendering and 1K–4K output choices for designed image routes.
Start with a 1K JPG draft, then choose 2K or 4K when the direction is approved.