Nano Banana

Nano Banana 2

Text-to-image and image edit via Google's Nano Banana 2 model, aka gemini-3.1-flash-image-preview. Fast, cheap, high-quality. Returns PNG.

When to use

• User asks for an image, picture, photo, thumbnail, cover art, poster, icon, etc.
• User provides an input image and wants it modified (background change, object removal, style swap)
• You need a quick visual for a blog post, README, demo, or slide

When NOT to use

• User wants a video — Nano Banana is image-only
• User wants a pixel-perfect UI mockup — use a design tool
• User needs a specific licensed stock photo — use a stock service

Setup

1. Get a Gemini API key: https://aistudio.google.com/app/apikey

2. Export it:

   export GEMINI_API_KEY="YOUR_KEY_HERE"
   # or
   export GOOGLE_API_KEY="YOUR_KEY_HERE"
   

3. Install uv if missing:

   command -v uv || curl -LsSf https://astral.sh/uv/install.sh | sh
   

First run auto-installs google-genai and pillow via uv's inline script deps.

Generate

uv run ~/.claude/skills/lich-skills/skills/nano-banana/scripts/generate_image.py \
    --prompt "A cinematic rainy Tokyo alley, neon reflections, 35mm, low angle" \
    --filename "tokyo-alley.png" \
    --resolution 1K

Output is saved to tokyo-alley.png in the current directory unless the filename is absolute.

Edit

Pass an input image with --input-image. The model sees the image + your instruction.

uv run .../generate_image.py \
    --prompt "Remove the crowd behind the subject. Keep framing and grading unchanged." \
    --filename "cleaned.png" \
    --input-image "./portrait.png" \
    --resolution 2K

When editing, the script auto-matches output resolution to the input's dimensions unless you explicitly pass --resolution.

Flags

Resolution mapping

Map loose user language to a resolution flag:

Prompt template (generation)

Subject: <subject>.
Style: <photo | illustration | 3D | painting | pixel art | ...>.
Composition: <shot type, angle>.
Lighting: <lighting description>.
Background: <background>.
Avoid: <what should NOT be in the image>.

Prompt template (editing)

Change only: <the specific edit>.
Keep subject, framing, lighting, palette, text, and overall style unchanged.

Being explicit about what to keep is the difference between a clean edit and a total re-render.

Environment

• Key lookup order: --api-key → GEMINI_API_KEY → GOOGLE_API_KEY
• Output format: PNG (RGBA converted to RGB on white background)
• Never prints the API key or image bytes to stdout

Fast checks

command -v uv
test -n "$GEMINI_API_KEY" || test -n "$GOOGLE_API_KEY"
# If editing:
test -f ./input.png

Common failures

Red flags

• Hardcoding the API key in any script or example
• Committing generated images to git without checking size
• Re-running the same prompt 10 times hoping for luck — instead, rewrite the prompt with the template above

Verification

• $GEMINI_API_KEY or $GOOGLE_API_KEY set in env, not in files
• A test generation produces a PNG at the expected path
• gitleaks detect --source . clean