Skip to main content
Text to Image

Nano Banana 2

Generate and edit images using Gemini 3.1 Flash Image Preview with optional reference images

View details

Inputs

Loading input fields...
Execution Steps

Loading workflow structure...

Loading curated examples...

Overview

Nano Banana 2 generates new images from a prompt or edits images you provide through source image URLs. Use it when you need fast campaign visuals, product-style variations, ad concepts, social creative, or source-aware edits from a plain-language brief.

Use cases

  • Draft several visual directions for a landing page, ad, or social post from one prompt.
  • Edit reference images by describing the change, style, composition, or visual treatment you want.
  • Generate multiple aspect-ratio or resolution variants for creative exploration.

Input tips

  • Write the subject, setting, visual style, composition, and intended use in the prompt.
  • Add image_urls only when you want the AI Tool to edit or reference existing images.
  • Provide 1 to 14 source image URLs; an empty image list is treated as prompt-only generation.
  • Choose 1 to 4 output images, an aspect ratio, and a 0.5K, 1K, 2K, or 4K resolution.
  • Use HTTPS image URLs that can be downloaded without login.

Expected output

The AI Tool returns one or more generated image URLs plus image metadata such as content type, file name, file size, width, and height when available, a short description of the generated or edited image set, and cost metadata. Results are handed off as downloadable image assets instead of a pasted data blob.

Caveats

  • Prompt wording strongly affects image quality, style, and brand fit.
  • Source image URLs must be publicly reachable image files; failed downloads can stop an edit run.
  • Generated images may need human review for accuracy, policy fit, and brand standards.
  • If the model returns no usable image data, the run fails instead of returning an empty gallery.
  • Large or high-resolution multi-image requests can take longer than a single 1K image.