The job search that runs on your machine. AI job application framework built on Claude Code: evaluate postings, tailor CVs, write cover letters, prep interviews. Fork it and own it.
Claude Code system for image prompt writing
Claude Banana helps Claude Code turn a short image idea into a detailed generation prompt. It does this with a prompt formula, a knowledge base of techniques and modes, reusable templates, and optional scripts for generating or editing images directly.
Builders who use Claude Code to turn rough image ideas into stronger prompts.
You can go from a vague image idea to a detailed prompt that is easier for image models to follow.
What it does
Prompt formula
Builds prompts from seven parts: subject, style, environment, lighting, action, camera, and texture.
Domain modes
Adapts prompts for cinema, product, portrait, fashion, UI/web, logo, landscape, abstract, and infographic images.
Creative techniques catalog
Includes 70+ techniques for styles like anime, crystal balls, frame breaks, sketch-photo mixes, tilt-shift cities, and voxel art.
Brand presets
Uses preset style files for repeatable colors, fonts, and mood across a set of images.
Prompt templates
Provides ready-to-use templates in `templates/examples/` for customizing common image prompt patterns.
Image generation scripts
Offers Python scripts to validate setup, generate images, edit images, and batch variations from the terminal.
How to get it
- 1Run
git clone https://github.com/Hainrixz/claude-banana.git cd claude-banana
- 2Run
claude
- 3Then just describe what you want in plain words
"Help me create a prompt for a cozy coffee shop with warm lighting and a cat sleeping on some books"
README
Claude Banana
You tell it what picture you want. It writes the perfect instructions to make that picture.
That's it. That's what Claude Banana does.
The Problem
You want AI to make you an image. You type "a cool sunset." You get... something okay. But not what you imagined.
Why? Because AI image generators need very specific instructions. The more detail you give them — lighting, camera angle, colors, mood, textures — the better the result. But writing those detailed instructions is hard and takes practice.
The Solution
Claude Banana is your creative assistant. You describe what you want in plain words, and it:
- Asks you a few quick questions to understand your vision
- Builds a detailed, optimized prompt using a proven formula
- Gives you a ready-to-copy prompt that gets amazing results from image generators
Think of it like having a professional photographer in your pocket who translates "I want a cool sunset" into a paragraph of perfect instructions that the AI actually understands.
How to Use It
Step 1: Get Claude Code
Install Claude Code (Anthropic's command-line tool).
Step 2: Download This Project
git clone https://github.com/Hainrixz/claude-banana.git
cd claude-banana
Step 3: Start Claude and Ask
claude
Then just describe what you want in plain words:
"Help me create a prompt for a cozy coffee shop with warm lighting and a cat sleeping on some books"
Claude automatically picks up the prompt-architect agent and knowledge base from this project. It will ask you a couple of questions (like "What's the mood?" or "Where will you use this image?"), and then give you a perfect prompt ready to paste into any image generator.
What Makes It Special
The 7-Piece Formula
Every prompt Claude Banana creates has 7 carefully balanced ingredients:
| Ingredient | What It Does | Example |
|---|---|---|
| Subject | Who or what is in the picture | "A gray tabby cat curled up on a stack of books" |
| Style | How it looks visually | "Warm watercolor illustration with soft edges" |
| Environment | Where it takes place | "Inside a small bookshop cafe with wooden shelves" |
| Lighting | How light falls on the scene | "Honey-gold afternoon light from a bay window" |
| Action | What's happening | "The cat is sleeping with one paw draped over a book spine" |
| Camera | How the picture is framed | "Eye-level close-up, slightly overhead angle" |
| Texture | What things feel like | "Worn leather book covers, soft fur, steam rising from a mug" |
9 Expertise Modes
The agent adjusts its formula depending on what kind of image you want:
- Cinema — Movie-like dramatic scenes
- Product — Clean, professional product photos
- Portrait — Beautiful photos of people
- Fashion — Magazine-style editorial looks
- UI/Web — App icons, web designs, interfaces
- Logo — Brand marks and symbols
- Landscape — Nature and environments
- Abstract — Artistic patterns and textures
- Infographic — Charts, data visuals, diagrams
70+ Creative Techniques
Want something special? The agent knows how to do things like:
- Turn photos into anime/Ghibli style
- Create miniature worlds inside crystal balls
- Make characters break out of picture frames
- Mix pencil sketches with real photos
- Build tiny tilt-shift cities
- Design voxel/pixel art icons
- And 60+ more creative effects
Works With Your Existing Prompts
Already have prompts from Midjourney, DALL-E, or Stable Diffusion? Paste them in and Claude Banana converts them to work better with Nano Banana Pro (Google's image generator).
Brand Presets
Making images for a brand? Load a preset with your brand colors, style, and mood so every image stays consistent.
What's Inside
claude-banana/
knowledge/ -- The brain: everything the agent knows about making great prompts
templates/examples/ -- 25 ready-to-use prompt templates you can customize
presets/ -- Brand style presets (colors, fonts, mood)
scripts/ -- Python tools for generating images directly
.claude/agents/ -- The agent itself
Optional: Generate Images Directly
If you have a Google AI API key (free from aistudio.google.com), you can generate images without leaving your terminal:
# Check your setup
python3 scripts/validate_setup.py
# Generate an image from a prompt
python3 scripts/generate.py --prompt "your optimized prompt here"
# Generate from a template with custom values
python3 scripts/generate.py --template templates/examples/cinematic-landscape.md \
--vars '{"location": "mountain lake", "time_of_day": "sunset"}'
# Edit an existing image
python3 scripts/edit.py --image photo.png --instruction "make it look like autumn"
# Generate a batch of variations
python3 scripts/batch.py --template templates/examples/product-showcase.md \
--variations my_products.json
Want to Contribute?
We'd love your help! You can:
- Share a prompt template — Open an issue with your favorite prompt
- Add a technique — Document a new creative technique
- Improve the knowledge — Make the reference guides even better
- Fix bugs — Submit a pull request
See the issue templates for easy submission forms.
License
MIT — free to use, modify, and share. See LICENSE.
Files in the repo
- .claude
- .github
- knowledge
- presets
- scripts
- templates
- .gitignore
- CLAUDE.md
- LICENSE
- README_ES.md
- README.md
Discussion (0)
Ask about usage, or say what you built with itSign in to join the discussion.
No comments yet. Be the first to say what this is good for.
More harnesses
A local multi-agent harness that works with your existing Claude Code, Codex subscriptions, allows you to run an office of agents
Open-source agentic workspace enterprises can make their own. Connect the systems you already run — 100+ integrations, MCP, chat tools, apps, browser, local files — with shared memory. Any agent (Claude Code, Codex), any model, or BYOK. Set up in clicks, not months. Local-first: your data never leaves your machines.
A realtime voice runtime that keeps Agents talking, working, and present. Real-time Voice Runtime for AI Agents
The long-horizon computer-use harness. Run AI agents across desktop apps and the CLI for extended periods while preserving task state and making reliable progress on complex workflows. Features fresh-context execution, durable verified state, independent auditing, recoverable progress, and native Claude Code / Codex / OpenClaw integration.
The one and only agent harness for complex codebases. Project memory, planning, execution, and verified completion inside Codex.