Skip to main content

Gemini 3.1 Flash integration

The faster variant of Nano Banana Pro, powered by Google’s Gemini 3.1 Flash model for rapid image generation with optional web search grounding.
Nano Banana Pro Flash (also known as Nano Banana 2 on fal.ai) is a text-to-image API powered by Google’s Gemini 3.1 Flash model. It generates images from natural language prompts with faster processing times than the standard Nano Banana Pro, making it suited for iterative creative workflows and high-volume applications. The model supports Google Search grounding, reference-guided generation with up to 3 images, 10 aspect ratios, and resolutions up to 4K.

​
Key capabilities

  • Fast generation: Powered by Gemini 3.1 Flash for shorter processing times compared to Nano Banana Pro
  • Google Search grounding: Enable use_google_search_tool for images that need real-world accuracy (places, brands, current events)
  • Reference image support: Up to 3 reference images (PNG, JPEG, WebP) for style and subject guidance
  • Flexible aspect ratios: 10 ratios including 1:1, 16:9, 9:16, 21:9, 2:3, 3:2, 4:3, 3:4, 5:4, 4:5
  • Resolution options: 1K (fast), 2K (balanced), 4K (best quality)
  • Prompt range: 2 to 3,000 characters for detailed scene descriptions
  • Async processing: Webhook notifications or polling for task completion

​
Use cases

  • Rapid prototyping: Fast iteration on visual concepts where speed matters more than maximum quality
  • Grounded generation: Images of real places, products, or events using Google Search for accuracy
  • Reference-guided creation: Style transfer and subject guidance from existing images
  • Marketing and social media: Quick generation of campaign visuals in various aspect ratios
  • High-volume pipelines: Batch generation where throughput is a priority

​
Generate images with Nano Banana Pro Flash

Create images by submitting a text prompt to the API. The service returns a task ID for async polling or webhook notification.

POST /v1/ai/text-to-image/nano-banana-pro-flash

Create a new image generation task

GET /v1/ai/text-to-image/nano-banana-pro-flash

List all Nano Banana Pro Flash tasks

GET /v1/ai/text-to-image/nano-banana-pro-flash/{task-id}

Get task status and results by ID

​
Parameters

​
Frequently Asked Questions

Nano Banana Pro Flash is a text-to-image API powered by Google’s Gemini 3.1 Flash model. You submit a text prompt (2-3,000 characters) via the API, optionally include up to 3 reference images, and receive a task ID immediately. Poll the task endpoint or configure a webhook to receive the generated image URL when processing completes.
When use_google_search_tool is set to true, the model uses Google Search to ground its generation with real-world information. This improves accuracy for prompts referencing real places, people, brands, or current events. The response includes groundingMetadata with source information when this feature is active.
Nano Banana Pro Flash accepts PNG, JPEG, and WebP reference images. Each reference image must be a publicly accessible URL. You can include up to 3 reference images, each with optional descriptive text and mime type.
Three resolution levels: 1K (fastest generation), 2K (balanced quality and speed), and 4K (best quality).
Rate limits depend on your subscription tier. See Rate Limits for current limits.
See the Pricing page for current rates and subscription options.
Nano Banana Pro Flash uses Google’s Gemini 3.1 Flash model optimized for speed, while Nano Banana Pro uses the standard model for maximum quality. Both share the same parameters, aspect ratios, and resolution options. Choose Flash when generation speed matters; choose Pro when you need the highest quality output. Flash defaults to 1K resolution while Pro defaults to 2K.

​
Best practices

  • Google Search grounding: Enable use_google_search_tool when your prompt references real-world entities (landmarks, brands, public figures) for more accurate results
  • Reference images: Use high-quality reference images with clear visual elements for best guided generation
  • Prompt writing: Be specific about subjects, composition, lighting, and style. The 3,000-character limit allows for detailed scene descriptions
  • Resolution selection: Use 1K for fastest iteration, 2K for balanced workflows, 4K for final production assets
  • Production integration: Use webhooks instead of polling for scalable applications
  • Error handling: Implement retry logic with exponential backoff for 503 errors
  • Mystic: Magnific’s proprietary image generation with LoRA style control
  • Flux Kontext Pro: Context-aware generation with image-guided editing
  • Seedream 4: High-quality text-to-image generation with advanced prompt understanding