Tools · Free Agent Skill · yeoudev
generate-nanobanana
AI image & video generation for your coding agent — powered by Google Gemini.
One command that routes the job to the right Gemini model — Nano Banana 2 Lite for cheap drafts, Nano Banana Pro for finals, Gemini Omni Flash for video — quotes the cost and waits for your go-ahead before anything paid runs, and logs the exact prompt beside every file it creates. Built for Claude Code, Antigravity, and Cursor.
Why This Skill
Built to never surprise you with a bill
One key, no markup
Google-only by design — one API key, one bill, no aggregator in the middle.
Cost gates every call
Quotes the model and price, then waits for your explicit go-ahead — image or video.
Real reference images
Loads your actual logo, face, or brand shots instead of describing them in words.
Nothing gets lost
Every output gets a sidecar JSON with the exact prompt, model, and cost beside it.
Models
One skill, four Gemini models
- Nano Banana 2 Lite — cheap draft images, ~$0.03 / image (1K)
- Nano Banana 2 — standard quality finals, $0.05–0.15 / image (0.5K–4K)
- Nano Banana Pro — up to 14 reference images, dense on-image text, $0.13–0.24 / image
- Gemini Omni Flash — video with synced audio, ~$0.10 / second (720p)
Prices are Google's published Gemini API pricing as documented in the repo — check the README for the current figures.
Installation
Up and running in a few minutes
- Get a Google AI Studio API key (or use Antigravity's zero-key built-in fallback).
- Run npx skills add AntonioCardenas/generate-nanobanana in your project.
- Set GEMINI_API_KEY in your shell profile or a project .env file.
- Restart your agent session if this is the first skill you've ever installed.
- Ask: "/generate a 16:9 thumbnail for my article, use refs/logo.png"
Install command
npx skills add AntonioCardenas/generate-nanobananaSupport the project
Saving you time (or money)?
It's free and open source — a star on GitHub helps other developers find it.
FAQ
Common questions
- Is generate-nanobanana free?
- The skill itself is free and open source under the MIT license. Gemini API usage is billed by Google per generation (a few cents per image, about $0.10/second for video) — the skill always quotes the cost and waits for your approval before any paid call runs.
- Which agents does it support?
- Claude Code, Antigravity, Antigravity CLI, Cursor, Codex, OpenCode, and any other agent that reads skill files. Install it with the npx skills CLI and target a specific agent with -a, or install it globally with -g.
- Does it guarantee reproducible output?
- No — Google's Gemini API doesn't expose a seed parameter, so a rerun is a fresh generation, not a pixel-identical repeat. Instead, every output logs its exact prompt, references, and response ID so you can reuse them precisely, or feed an approved image back in as a reference for a matching series.
- What does it access on my machine?
- Only generativelanguage.googleapis.com over HTTPS, carrying your prompt, references, and API key — no other endpoint. It reads GEMINI_API_KEY from your environment or a .env file you create yourself, and writes files only inside a generations/ folder in your current project.
Ready to generate on-brand images?
Free, open source, cost-gated, and logged. No aggregator, just Gemini.
View on GitHub →