Nano Banana Pro Image Generation & Editing

Quick Start

Generate:

uv run ~/.codex/skills/nano-banana-pro/scripts/generate_image.py --prompt "A serene Japanese garden" --filename "2025-11-23-14-23-05-japanese-garden.png" --resolution 1K

Edit:

uv run ~/.codex/skills/nano-banana-pro/scripts/generate_image.py --prompt "make the sky dramatic" --filename "2025-11-23-14-25-30-dramatic-sky.png" --input-image "original.png" --resolution 2K

Compose:

uv run ~/.codex/skills/nano-banana-pro/scripts/compose_images.py "Create a group photo" group.png person1.png person2.png

Multi-turn chat:

uv run ~/.codex/skills/nano-banana-pro/scripts/multi_turn_chat.py --model gemini-3-pro-image-preview --output-dir .

Scripts

• generate_image.py: text-to-image + edit via --input-image, resolution 1K/2K/4K, optional --aspect, auto-detect resolution on edit.
• edit_image.py: explicit edit CLI (input, instruction, output).
• compose_images.py: combine up to 14 reference images.
• multi_turn_chat.py: interactive refine session (/save, /load, /clear, /quit).
• gemini_images.py: Python helper class.

Model

• Default: gemini-3-pro-image-preview.
• Only use other models if user asks.

Resolution + Aspect

• Resolution: 1K default, 2K, 4K.
• Aspect ratios (where supported): 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9.
• Map words: "1080/low/1K"->1K, "2K/2048/medium"->2K, "4K/high/ultra"->4K.

Workflow

• Draft 1K, iterate small prompt diffs, final 4K.
• For edits, keep same --input-image until final.

API Key

• GEMINI_API_KEY required.
• generate_image.py also supports --api-key.

Filenames

• Pattern: yyyy-mm-dd-hh-mm-ss-name.png.
• Lowercase, hyphenated, 1-5 words.

Output

• Run from user cwd so files save there.
• generate_image.py converts output to PNG.
• Do not open/read images; report saved path.

Preflight

• command -v uv
• test -n "$GEMINI_API_KEY"
• If editing: test -f "path/to/input.png"

Common Failures

• No API key.
• Input image missing/unreadable.
• Quota/permission/403 errors.