name: nanobanana description: Use this skill to generate or edit images using Google's latest AI models.

Nano Banana Skill

This skill allows you to generate and edit images using Google's Gemini models via the nanobanana.js script. It supports text-to-image and multimodal (image + text) prompts.

Prerequisites

Ensure the GEMINI_API_KEY environment variable is set.

Usage

node nanobanana.js [options] <prompt> [image_path] ...

Options

• --model <name>: Specifies the Gemini model to use. (Default: gemini-2.5-flash-image)
• --output, -o <path>: Output filename or directory. (Default: ./nanobanana-outputs)
• --aspectRatio, --ar <ratio>: Sets the aspect ratio of the generated image (e.g., 16:9, 1:1, 4:3, 3:4).
• --count <number>: Number of images to generate.
• --seed <number>: Integer seed for reproducible results.
• --imageSize <size>: Requires gemini-3-pro-image-preview. Sets specific image dimensions (e.g., 1024x1024).
• --googleSearch: Requires gemini-3-pro-image-preview. Enables Google Search grounding for generation. (Default: false)

Constraints

• Model Capabilities: The --imageSize and --googleSearch flags are only compatible with the gemini-3-pro-image-preview model. Using them with other models may result in errors or them being ignored.

Examples

1. Basic Text-to-Image

Generate a landscape using the default model (gemini-2.5-flash-image).

node nanobanana.js "A futuristic city at sunset, synthwave style"

2. Image Variation (Image-to-Image)

Generate a variation of an existing image.

node nanobanana.js "A watercolor painting of this scene" ./images/01_Evan_Plays_Drums.png

3. Intermixed Text & Images (Virtual Try-On)

Provide multiple text and image inputs in sequence to guide the generation.

node nanobanana.js "Show this person:" ./person.jpeg "wearing this shirt:" ./shirt.png

4. Advanced Generation with Preview Model

Use advanced features like imageSize and googleSearch with the preview model.

node nanobanana.js --model gemini-3-pro-image-preview --imageSize 1024x1024 --googleSearch "A realistic image of the latest electric car model"

5. Specific Aspect Ratio, Count and Output

Generate 3 portraits with a 9:16 aspect ratio and save them to a specific directory.

node nanobanana.js --ar 9:16 --count 3 --output ./my-portraits/ "Portrait of a cypherpunk character"

Output

Generated images are saved to the ./nanobanana-outputs directory with unique filenames (e.g., image-uuid.png).