Nano Banana Pro Image Generation & Editing

Generate and edit images using Google's Nano Banana Pro (Gemini 3 Pro Image) — the premium AI image generation model designed for professional asset production, utilizing advanced reasoning ("Thinking") to follow complex instructions and render high-fidelity text in images.

Nano Banana Pro excels at infographics, menus, diagrams, marketing assets, and any task requiring precise text rendering and complex multi-object composition.

Data usage note: This skill sends text prompts and image URLs/data to Atlas Cloud API for image generation. No data is stored locally beyond the downloaded output files.

Security note: API keys are read exclusively from environment variables (ATLASCLOUD_API_KEY) and passed via HTTP headers — never embedded in URL query strings or command arguments. All user-provided text (prompts, file paths) must be passed through JSON request bodies to prevent shell injection. When constructing curl commands, always use a JSON payload (-d '{...}') rather than string interpolation in the shell. File paths should be validated before use. The skill does not execute any user-provided code — it only sends structured API requests and downloads output files.

Nano Banana Pro vs Nano Banana 2

Choose Nano Banana Pro when quality and text precision matter. Choose Nano Banana 2 when speed and cost matter.

Pricing

Atlas Cloud uses flat-rate pricing regardless of resolution.

Available Models

Atlas Cloud API

Setup

1. Sign up at https://www.atlascloud.ai
2. Console → API Keys → Create new key
3. Set env: export ATLASCLOUD_API_KEY="your-key"

Parameters

Text-to-Image:

Image Editing — same as above plus:

Workflow: Submit → Poll → Download

# Step 1: Submit
curl -s -X POST "https://api.atlascloud.ai/api/v1/model/generateImage" \
  -H "Authorization: Bearer $ATLASCLOUD_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "google/nano-banana-pro/text-to-image",
    "prompt": "A professional infographic showing quarterly revenue growth with bar charts and annotations",
    "aspect_ratio": "16:9",
    "resolution": "2k"
  }'
# Returns: { "code": 200, "data": { "id": "prediction-id" } }

# Step 2: Poll (every 3-5 seconds until "completed" or "succeeded")
curl -s "https://api.atlascloud.ai/api/v1/model/prediction/{prediction-id}" \
  -H "Authorization: Bearer $ATLASCLOUD_API_KEY"
# Returns: { "code": 200, "data": { "status": "completed", "outputs": ["https://...url..."] } }

# Step 3: Download
curl -o output.png "IMAGE_URL_FROM_OUTPUTS"

Image editing example:

curl -s -X POST "https://api.atlascloud.ai/api/v1/model/generateImage" \
  -H "Authorization: Bearer $ATLASCLOUD_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "google/nano-banana-pro/edit",
    "prompt": "Replace the text on the sign with: Grand Opening Sale — 50% Off",
    "images": ["https://example.com/storefront.jpg"],
    "resolution": "2k"
  }'

Polling logic:

• processing / starting / running → wait 3-5s, retry (Pro model may take longer than Nano Banana 2 due to Thinking mode)
• completed / succeeded → done, get URL from data.outputs[]
• failed → error, read data.error

Atlas Cloud MCP Tools (if available)

If the Atlas Cloud MCP server is configured, use built-in tools:

atlas_quick_generate(model_keyword="nano banana pro", type="Image", prompt="...")
atlas_generate_image(model="google/nano-banana-pro/text-to-image", params={...})
atlas_get_prediction(prediction_id="...")

Implementation Guide

1. Check API key: Ensure ATLASCLOUD_API_KEY is set. If not, ask the user to sign up at https://www.atlascloud.ai and set export ATLASCLOUD_API_KEY="your-key".

2. Extract parameters:

   • Prompt: the image description — Nano Banana Pro handles complex, detailed prompts well
   • Aspect ratio: infer from context (infographic→3:4 or 9:16, banner→16:9, menu→3:4, social post→1:1)
   • Resolution: default 1k, use 2k/4k for professional output
   • For editing: identify source image URL(s) or local file path

3. Choose model tier:

   • Standard for production use
   • Developer if user wants to save costs or is experimenting

4. Sanitize inputs: Ensure user-provided prompts and file paths do not contain shell metacharacters. Always pass prompts inside JSON payloads (never via shell interpolation). Validate that image file paths exist and are readable before encoding.

5. Execute: POST to Atlas Cloud generateImage API → poll prediction (may take 10-30s due to Thinking mode) → download result

6. Present result: show file path, offer to open

Prompt Tips for Nano Banana Pro

Nano Banana Pro excels at understanding complex, structured prompts. Take advantage of its Thinking mode:

• Text in images: Include exact text in quotes — Pro renders text with high fidelity. Example: "A cafe chalkboard menu reading: 'Today's Special — Matcha Latte $5.50'"
• Infographics: Describe data, layout, and annotations. Example: "An infographic showing 3 steps of coffee brewing with numbered icons and captions"
• Marketing assets: Specify brand colors, text placement, and style. Example: "A product banner with dark background, gold accents, text 'Limited Edition' top-center"
• Complex compositions: Describe spatial relationships and multiple objects. Example: "A still life with a ceramic vase left-center, three oranges arranged in front, and a linen cloth draped over the table edge"
• Style: "photorealistic", "editorial illustration", "minimalist flat design", "watercolor"
• Lighting: "studio lighting", "natural window light", "dramatic chiaroscuro"