A powerful command-line tool for generating and editing images using Google's Gemini AI. Features both an interactive terminal UI and a scriptable CLI interface.
Features • Installation • Quick Start • CLI Reference • TUI Guide
| Feature | Description |
|---|---|
| Dual Interface | Interactive TUI for exploration, CLI for scripting and automation |
| Image Generation | Create images from text prompts with Gemini AI |
| Image Editing | Transform existing images using reference-based editing |
| Style Analysis | Extract style descriptions from images with describe command |
| Parallel Processing | Generate up to 20 images simultaneously |
| Flexible Output | Control aspect ratio (1:1, 16:9, 9:16, 4:3, 3:4) and size (1K, 2K, 4K) |
| Google Search Grounding | Enhance prompts with real-time web search context |
| Path Autocomplete | Tab completion for file paths in TUI (supports ~ expansion) |
Download the latest release for your platform:
| Platform | Architecture | Download |
|---|---|---|
| macOS | Apple Silicon (M1/M2/M3) | banana-darwin-arm64 |
| macOS | Intel | banana-darwin-amd64 |
| Linux | x64 | banana-linux-amd64 |
| Linux | ARM64 | banana-linux-arm64 |
| Windows | x64 | banana-windows-amd64.exe |
| Windows | ARM64 | banana-windows-arm64.exe |
After downloading, make it executable (macOS/Linux):
chmod +x banana-darwin-arm64
mv banana-darwin-arm64 /usr/local/bin/bananago install github.com/AhmedAburady/banana-cli/cmd/banana@latestgit clone https://github.com/AhmedAburady/banana-cli.git
cd banana-cli
go build -o banana ./cmd/bananaGet a free Gemini API key from Google AI Studio.
Option A: Save to config file (Recommended)
banana config set-key YOUR_API_KEYOption B: Environment variable
# Add to your shell profile (~/.bashrc, ~/.zshrc, etc.)
export GEMINI_API_KEY="your_api_key_here"Option C: Just run it
# CLI will prompt you to enter and save your API key
banana -p "a sunset"
# TUI will show an API key input screen
bananaUsing TUI (Interactive):
bananaUsing CLI (One-liner):
banana -p "a cyberpunk city at night with neon lights"The CLI mode allows you to generate or edit images directly from the command line, perfect for scripting and automation.
banana [flags]
banana describe [flags]
banana config <command>
Running banana without flags opens the interactive TUI.
| Flag | Long Form | Type | Description | Default |
|---|---|---|---|---|
-p |
string | Prompt - The text description for image generation | required | |
-o |
string | Output - Directory to save generated images | . (current) |
|
-n |
int | Number - How many images to generate (1-20) | 1 |
|
-ar |
string | Aspect Ratio - Image dimensions ratio | Auto |
|
-s |
string | Size - Output resolution | 1K |
|
-g |
bool | Grounding - Enable Google Search grounding | false |
|
-i |
string | Input - Reference image/folder for edit mode | none | |
-v |
--version |
Show version | ||
-vertex |
bool | Use Vertex AI instead of Gemini API | false |
|
--help |
Show help message |
Manage your API key configuration:
banana config set-key <KEY> # Save your Gemini API key
banana config show # Show current configuration (key is masked)
banana config path # Show config file locationThe config file is stored at ~/.config/banana/config.json.
Analyze images and extract style descriptions:
banana describe -i <image-or-folder> [flags]| Flag | Description | Default |
|---|---|---|
-i |
Input image or folder (required) | - |
-o |
Output file path | stdout |
-p |
Custom prompt (overrides default) | - |
-a |
Additional context (prepended to default) | - |
-json |
Output as structured JSON | false |
| Value | Use Case |
|---|---|
Auto |
Default - Let Gemini decide the best ratio |
1:1 |
Square - Social media posts, profile pictures |
16:9 |
Landscape - Desktop wallpapers, YouTube thumbnails |
9:16 |
Portrait - Phone wallpapers, Instagram stories |
4:3 |
Classic - Presentations, traditional photos |
3:4 |
Portrait Classic - Portraits, posters |
2:3 |
Portrait - Standard photo print ratio |
3:2 |
Landscape - DSLR camera ratio |
5:4 |
Near-square - Medium format photos |
4:5 |
Portrait - Instagram portrait posts |
21:9 |
Ultra-wide - Cinematic, ultrawide monitors |
| Value | Resolution | Best For |
|---|---|---|
1K |
~1024px | Quick previews, web use |
2K |
~2048px | High-quality prints, detailed work |
4K |
~4096px | Maximum quality, large prints |
Generate images from text prompts:
# Simple generation - creates 1 image in current directory
banana -p "a mountain landscape at sunset"
# Multiple images with custom output folder
banana -p "abstract geometric patterns" -n 5 -o ./my-patterns
# Widescreen wallpaper in 4K
banana -p "northern lights over a snowy forest" -ar 16:9 -s 4K -o ~/Wallpapers
# Phone wallpaper
banana -p "minimalist gradient with soft colors" -ar 9:16 -o ./phone-wallpapers
# With Google Search grounding for current/real-world topics
banana -p "the latest Tesla Cybertruck design" -gTransform existing images using the -i flag:
# Edit a single image
banana -i ./photo.jpg -p "convert to watercolor painting style"
# Use multiple reference images from a folder
banana -i ./reference-images/ -p "create a pattern inspired by these designs" -n 3
# Style transfer
banana -i ./portrait.png -p "transform into anime art style" -o ./anime-versions
# Add effects
banana -i ./landscape.jpg -p "add dramatic storm clouds and lightning"Extract style descriptions from images:
# Plain text style description
banana describe -i photo.jpg
# Analyze folder of style references (unified description)
banana describe -i ./reference_images/
# Add style context to guide analysis
banana describe -i image.png -a "2D flat vector art"
# Structured JSON output
banana describe -i photo.jpg -json -o style.json⠋ Generating 3 image(s)...
✓ generated_1_20260123_143052.png
✓ generated_2_20260123_143053.png
✓ generated_3_20260123_143054.png
Done: 3 success, 0 failed (12.4s)
Output: /Users/ahmed/my-images
The Terminal User Interface provides an interactive experience for image generation.
bananaFrom the main menu, choose between:
- Generate Image - Create new images from text prompts
- Edit Image - Transform existing images with AI
| Field | Description |
|---|---|
| Output Folder | Where to save images (supports ~ and tab completion) |
| Number of Images | 1-20 images generated in parallel |
| Prompt | Your image description |
| Aspect Ratio | Select from 1:1, 16:9, 9:16, 4:3, 3:4 |
| Image Size | Select from 1K, 2K, 4K |
| Grounding | ON/OFF - Enable Google Search grounding |
Same as Generate, plus:
| Field | Description |
|---|---|
| Reference Path | Image file or folder containing reference images |
| Key | Action |
|---|---|
↑ ↓ |
Navigate between form fields |
← → |
Cycle through options (aspect ratio, size, grounding) |
Tab |
Accept path autocomplete suggestion |
Ctrl+N |
Insert newline in prompt field |
Ctrl+S |
Submit form and start generation |
Esc |
Go back to previous screen |
Enter |
Select menu item |
q |
Quit from main menu |
Ctrl+C |
Force quit |
The TUI supports intelligent path autocomplete:
- Start typing a path and press
Tabto see suggestions - Use
~for home directory (e.g.,~/Pictures) - Works for both output folders and reference images
| Format | Extensions |
|---|---|
| JPEG | .jpg, .jpeg |
| PNG | .png |
| GIF | .gif |
| WebP | .webp |
All generated images are saved as PNG format.
┌─────────────────────────────────────────────────────────────┐
│ BANANA CLI │
├─────────────────────────────────────────────────────────────┤
│ │
│ ┌─────────┐ ┌─────────────┐ ┌─────────┐ │
│ │ TUI │────────▶│ │───────▶│ Save │ │
│ └─────────┘ │ Gemini │ │ PNG │ │
│ │ API │ │ Files │ │
│ ┌─────────┐ │ │ └─────────┘ │
│ │ CLI │────────▶│ (Parallel) │ │
│ └─────────┘ └─────────────┘ │
│ │
└─────────────────────────────────────────────────────────────┘
- Input: Prompt text + optional reference images
- Processing: Parallel API calls to Gemini (up to 20 concurrent)
- Output: PNG images saved to specified folder
BANANA CLI looks for your API key in the following order (first found wins):
| Priority | Source | Description |
|---|---|---|
| 1 | GEMINI_API_KEY |
Environment variable (highest priority) |
| 2 | GOOGLE_API_KEY |
Alternative environment variable |
| 3 | Config file | ~/.config/banana/config.json |
This allows you to:
- Use environment variables to temporarily override the saved key
- Keep a default key in the config file for convenience
- Use different keys for different projects via env vars
Instead of using a Gemini API key, you can use Vertex AI with Google Cloud authentication. This is useful if you have a GCP project with better quotas or enterprise features.
- A Google Cloud project with the Vertex AI API enabled
- A service account with the Vertex AI User role (or equivalent)
- Google Cloud CLI (
gcloud) installed
Step 1: Authenticate with Google Cloud
# Login with your Google account
gcloud auth application-default login
# Or use a service account
gcloud auth activate-service-account --key-file=your-service-account.json
export GOOGLE_APPLICATION_CREDENTIALS="path/to/your-service-account.json"Step 2: Configure your GCP project
Option A: Save to config file (recommended)
banana config set-project your-project-id
banana config set-location global # optional, defaults to globalOption B: Use environment variables
export GOOGLE_CLOUD_PROJECT="your-project-id"
export GOOGLE_CLOUD_LOCATION="global" # optionalStep 3: Use the -vertex flag
banana -p "a beautiful sunset" -vertex
banana -p "cyberpunk city" -n 5 -vertex -ar 16:9Settings are loaded in this order (first found wins):
| Setting | Env Variable | Config Command | Default |
|---|---|---|---|
| GCP Project | GOOGLE_CLOUD_PROJECT |
banana config set-project |
- |
| GCP Location | GOOGLE_CLOUD_LOCATION |
banana config set-location |
global |
- No API key exposure - Uses Google Cloud IAM authentication
- Enterprise quotas - Higher rate limits based on your GCP project
- VPC Service Controls - Network security features
- Audit logging - Cloud Audit Logs integration
No API key configured:
# Easiest: save to config
banana config set-key YOUR_API_KEY
# Or use environment variable
export GEMINI_API_KEY="your_key_here"Check current configuration:
banana config show # Shows masked key
banana config path # Shows config file locationWhen using -i with a folder, ensure it contains supported image formats (.jpg, .jpeg, .png, .gif, .webp).
If generating many images, you may hit rate limits. The tool handles this gracefully - failed images will show error messages while successful ones are saved.
MIT License - see LICENSE for details.
Contributions are welcome! Feel free to:
- Report bugs
- Suggest features
- Submit pull requests
Made with Go and Gemini AI