WaveSpeedAI Nano Banana 2 Image Generation/Editing

Generate and edit images using Google's Nano Banana 2 model via the WaveSpeed AI platform. Supports both text-to-image generation and natural-language image editing with up to 14 input images.

Setup

Install the open-source CLI once and sign in; the CLI stores the key, so never ask the user to paste an API key into the chat:

npm install -g @wavespeed/cli
wavespeed login          # opens https://wavespeed.ai/accesskey and stores the key
wavespeed status         # confirms you are signed in

For CI or one-off shells, WAVESPEED_API_KEY in the environment also works.

Prefer MCP tools over shell commands? The same platform is exposed by @wavespeed/mcp (npx -y @wavespeed/mcp; tools search_models, get_model_schema, get_price, upload_file, run_model, get_prediction). It shares the CLI's stored login. Every example below maps one-to-one onto run_model with the same model id and input fields.

Quick Start

Text-to-Image

OUTPUT_URL=$(wavespeed run google/nano-banana-2/text-to-image \
  -p "A serene Japanese garden with cherry blossoms, watercolor style" \
  --json | jq -r '.outputs[0]')

Image Editing

The images parameter accepts an array of image URLs. For local files use the @path marker; the CLI uploads them for you.

# Upload a local image to get a URL
OUTPUT_URL=$(wavespeed run google/nano-banana-2/edit \
  -i images='["@./photo.png"]' \
  -p "Replace the sky with a dramatic sunset" \
  --json | jq -r '.outputs[0]')

Existing URLs work as-is:

OUTPUT_URL=$(wavespeed run google/nano-banana-2/edit \
  -i images='["https://example.com/photo.jpg"]' \
  -p "Replace the sky with a dramatic sunset" \
  --json | jq -r '.outputs[0]')

API Endpoints

Text-to-Image

Model ID: google/nano-banana-2/text-to-image

Generate images from text prompts.

Parameters

Example

OUTPUT_URL=$(wavespeed run google/nano-banana-2/text-to-image \
  -p "A red vintage Porsche 911 on a winding mountain road at golden hour, photorealistic" \
  -i aspect_ratio="16:9" \
  -i resolution="2k" \
  -i output_format="png" \
  --json | jq -r '.outputs[0]')

Image Editing

Model ID: google/nano-banana-2/edit

Edit existing images using natural language prompts. Supports up to 14 input images.

Parameters

Example

# Upload local images first, or use existing URLs
OUTPUT_URL=$(wavespeed run google/nano-banana-2/edit \
  -i images='["@./living-room.png"]' \
  -p "Change the wall color to warm terracotta and add indoor plants" \
  -i aspect_ratio="16:9" \
  -i resolution="4k" \
  -i output_format="png" \
  --json | jq -r '.outputs[0]')

Multi-Image Editing

# Upload multiple local images
OUTPUT_URL=$(wavespeed run google/nano-banana-2/edit \
  -i images='["@./face.png", "@./hairstyle.png"]' \
  -p "Apply the hairstyle from the second image to the person in the first image" \
  --json | jq -r '.outputs[0]')

Aspect Ratio Options

Resolution and Pricing

Prompt Tips

• Be specific and descriptive: "A red vintage Porsche 911 on a winding mountain road at golden hour" vs "a car"
• Include style keywords: "digital art", "oil painting", "photorealistic", "watercolor", "cinematic"
• For edits, clearly describe the desired change: "Replace the sky with a dramatic sunset"
• For multi-image edits, reference images by position: "Apply the style from the second image to the first image"
• Leverage multilingual text rendering: the model supports on-image text with automatic translation
• Use camera-style control language: "shallow depth of field", "wide angle", "top-down view"

CLI tips

# Inspect the live input schema before running (fields, enums, defaults)
wavespeed run google/nano-banana-2/text-to-image -h

# Quote the price first
wavespeed price google/nano-banana-2/text-to-image -p "..." -i key=value

# Save outputs to disk instead of only printing URLs
wavespeed run google/nano-banana-2/text-to-image -p "..." --json --download "./out/{index}.{ext}"

# Local files: prefix the path with @ and the CLI uploads it and passes the hosted URL
wavespeed run google/nano-banana-2/text-to-image -i <field>=@./local-file.png --json

# Recover a result if the run was interrupted (the id is in the --json output)
wavespeed show <id>

run --json prints { id, model, prompt, outputs: [url, ...], saved: [path, ...], elapsed_ms, raw }. Read outputs[0] for the result URL.

Security constraints

• Never ask for the key in chat: wavespeed login handles auth; if wavespeed status says signed out, ask the user to run it.
• Local files only via @: bare paths are passed through untouched and the model will reject them. Only @-prefixed values upload.
• No arbitrary URL loading: only pass media URLs the user provided or that came back from a previous run.
• Input validation: only pass parameters documented above; confirm with wavespeed run <model> -h when unsure.