Overview

Nano Banana Pro lets you generate or edit images using the Gemini 3 Pro Image engine via a bundled script. It supports single prompts, single-image edits, and multi-image compositions up to 14 inputs, with output paths and optional timestamped filenames. Requires a GEMINI_API_KEY and offers 1K/2K/4K resolutions for flexible workflows.

How This Skill Works

You run the bundled generate_image.py script through uv, supplying a prompt and an output filename, plus an optional input image and resolution. For edits, include -i with the source image; for composite scenes, supply multiple -i inputs (up to 14). The script prints a MEDIA line for Moltbot, and you should report only the saved path—no image readback from the tool.

When to Use It

• You want to generate a new image from a detailed description.
• You need to edit an existing image using text prompts or style changes.
• You want to create a multi-image composition by combining up to 14 inputs into one scene.
• You must output a file with a timestamped filename for easy versioning.
• You are integrating image tasks into a workflow that uses GEMINI_API_KEY for authentication.

Quick Start

1. Step 1: Ensure GEMINI_API_KEY is set in your environment or in ~/.clawdbot/moltbot.json.
2. Step 2: Run a simple generate command, e.g.: uv run {baseDir}/scripts/generate_image.py --prompt "your image description" --filename "output.png" --resolution 1K
3. Step 3: If editing or composing, add -i input.png or multiple -i inputs and check the saved path reported by the script (do not expect the image to be read back).

Best Practices

• Start with a clear, descriptive prompt to reduce iteration time.
• Use 1K for quick drafts; switch to 2K or 4K only when finalizing.
• When editing, provide a precise -i input and specific edit instructions.
• Name outputs with timestamps (yyyy-mm-dd-hh-mm-ss-name.png) to avoid overwrites.
• Set GEMINI_API_KEY in the environment or in moltbot.json before running.

Example Use Cases

• Generate a futuristic cityscape at sunset with a focus on neon lighting (1K).
• Edit a portrait to enhance lighting and add a cyberpunk aesthetic (use -i input.png).
• Combine product shots of a smartwatch into a single marketing collage (up to 14 inputs).
• Create a large-scene fantasy battle by layering multiple input images (2K/4K).
• Produce a series of variant textures for a game asset pack by altering prompts.

Frequently Asked Questions