Nano Banana 2 Skill
github.com/kingbootoshi/nano-banana-2-skill- Category
- AI Tools
- Rank
- No. 1128Tools index
- Pricing
- Open Source
- Type
- TOOL
- Use case
- Design & Media
- Interfaces
- CLI · Agent Skill / Plugin
- Builder
- @kingbootoshi
- GitHub
- 414 stars
- Date
About
AI-powered CLI tool for generating images using Google's Gemini models with support for transparent backgrounds, style transfer, and reference images. Also works as a Claude Code plugin for seamless AI-assisted image generation workflows.
What it does
Nano Banana 2 turns a text prompt and command-line options into a Gemini image request, then writes the result locally. You can choose resolution, aspect ratio, model, output location, and multiple source images. An optional chroma-key pipeline produces cropped PNG assets, while local history records estimated generation costs.
Why it's ranked here
The tool has a focused, practical shape: one global command covers generation, image transformation, model selection, output sizing, and spending summaries. Its transparency workflow adds useful post-processing beyond a thin API wrapper. The trade is operational weight, since setup requires Bun, a Gemini key, and two more programs for transparent output.
What's good
The command surface supports resolutions from 512 through 4K, fourteen aspect ratios in the source, repeated reference inputs, custom model identifiers, and explicit output directories. API-key lookup supports command-line, environment, project, repository, and user-level configuration. Transparency samples a corner patch, detects the dominant key color, removes spill, trims padding, and cleans its temporary image.
Tradeoffs
Transparent output depends on FFmpeg and ImageMagick, and the documented installation command targets Homebrew users. Exact dimensions use a blank final reference rather than a direct width and height option. Cost records are local estimates, not provider billing data. Unrecognized model identifiers fall back to the default Flash pricing table, so their estimates may be inaccurate.
How to use it well
Use it for developers and designers producing UI concepts, marketing art, sprites, logos, or video elements from a terminal or Claude Code session. Start with low-resolution Flash drafts, introduce references for editing or style direction, then raise resolution or select Pro for final candidates. It does not expose a library or server API in the supplied package.
Technical notes+
package.json defines an ESM Bun package whose nano-banana binary points to src/cli.ts, with @google/genai as its direct dependency. src/cli.ts parses CLI arguments, resolves model aliases and API keys, base64-encodes references, persists cost entries under the user configuration directory, and spawns FFmpeg plus ImageMagick for chroma-key processing. .claude-plugin/marketplace.json registers the plugin source, while plugins/nano-banana/.claude-plugin/plugin.json supplies plugin metadata and plugins/nano-banana/skills/nano-banana/SKILL.md defines setup, triggers, workflows, and option guidance. bun.lock pins the dependency graph.
Observed
- License
- MIT
- Primary language
- TypeScript
- Runtime and package manager
- Bun
- Install surface
- Repository clone, Bun dependency installation, and global Bun link
- Interfaces
- Command-line executable and Claude Code plugin skill
- Direct dependency
- Google Gen AI JavaScript SDK
- Optional processing dependencies
- FFmpeg and ImageMagick are required for transparent mode
Read from README.md, package.json, src/cli.ts, LICENSE, bun.lock, .env.example, .claude-plugin/marketplace.json, plugins/nano-banana/.claude-plugin/plugin.json, plugins/nano-banana/skills/nano-banana/SKILL.md.
What it can do
Generate images from text prompts using AI
Text description/prompt → Generated image
Create images with transparent backgrounds
Text prompt with transparency requirements → Image with transparent/green screen background
Generate images at multiple resolutions
Text prompt and resolution specification (512px to 4K) → Image at specified resolution
Apply style transfer to generated images
Text prompt and style reference → Stylized image matching reference style
Generate images using reference images
Text prompt and reference image file → Generated image based on reference
Track image generation costs
Image generation requests → Cost tracking data and usage statistics
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.