peter119lee
ea11be980c
The publish job was gated on repository_owner == 'peter119lee', which the account rename made permanently false - every future registry publish would have been silently skipped. Gate on the repository fork flag instead, point repo links at the new slug (the registry PublisherId is a registry-side id and stays peter119lee), and bump the patch version so the registry listing refreshes its repository link.
🍌 ComfyUI-BananaForge
Gemini-Powered Image Generation for ComfyUI
Text-to-Image, Image-to-Image, and Batch Processing
✨ Features
- 🎨 Text → Image: Generate images from text prompts
- 🖼️ Image → Image: Edit, transform, and style transfer
- 📦 Batch Processing: Multiple prompts → multiple images
- ⏱️ Rate Limit Handling: Built-in wait timers
- 🔑 Flexible API Config: Direct key or .env file
📥 Installation
ComfyUI Manager (Recommended)
Search for "BananaForge" in ComfyUI Manager and install.
Manual Installation
cd ComfyUI/custom_nodes
git clone https://github.com/Rinne414/ComfyUI-BananaForge.git
pip install -r ComfyUI-BananaForge/requirements.txt
Restart ComfyUI after installation.
🚀 Quick Start
- Get an API Key from Google AI Studio
- Find nodes under "Banana Forge" category in ComfyUI
- Paste your API key
- Enter a prompt and run!
📋 Nodes Overview
🍌 BananaForge Text → Image
Purpose: Generate NEW images from text descriptions.
| Input | Description |
|---|---|
api_key |
Your Gemini API key |
model_name |
Default: gemini-3-pro-image-preview |
prompt |
Text description of desired image |
aspect_ratio |
1:1, 16:9, 9:16, 4:3, 3:4 |
image_size |
default, 2K, 4K |
wait_seconds |
Delay before API call (rate limit protection) |
Output: Single image
🖼️ BananaForge Image → Image
Purpose: Edit or transform existing images.
| Input | Description |
|---|---|
image_1 |
Primary input image (required) |
image_2 to image_6 |
Additional reference images (optional) |
prompt |
Editing instructions |
match_input_aspect |
Auto-match output to input image ratio |
Output: Edited image
Use Cases:
- Style transfer
- Image editing
- Character consistency (multiple refs)
- Image fusion
🍌 BananaForge Batch
Purpose: Generate multiple images from multiple prompts.
| Input | Description |
|---|---|
prompts |
One prompt per line (multi-line text) |
reference_image |
Optional: for batch i2i mode |
wait_between |
Seconds between API calls (default: 2s) |
Output:
IMAGES: Batch tensor (all generated images)COUNT: Number of successful generations
Example Prompts:
A girl sitting on a chair
A girl eating an apple
A girl standing by the window
→ Generates 3 separate images
⚙️ Configuration
Option 1: Direct API Key
Paste your key directly in the api_key field.
Option 2: Environment File (.env)
cp .env.example .env
Edit .env:
API_KEY=your_api_key_here
API_URL=https://generativelanguage.googleapis.com/v1beta
Set use_env_file to yes in the node.
🎯 Supported Models
| Model | Max Images | Sizes | Notes |
|---|---|---|---|
gemini-2.5-flash-image |
3 | default | Fast, free tier friendly |
gemini-3-pro-image-preview |
14 | 2K, 4K | Best quality, supports larger sizes |
💡 Tips
Avoid Rate Limits
- Use
wait_seconds(single nodes) orwait_between(batch) - Recommended: 2-5 seconds between calls
Better Results
- Be specific in prompts
- For i2i: describe desired changes clearly
- For batch: each line is separate, be complete in each
Third-Party APIs
Works with OpenAI-compatible APIs. Just change api_url to your provider.
🐛 Troubleshooting
| Error | Solution |
|---|---|
| "API key cannot be empty" | Add your key or check .env file |
| "Content blocked" | Modify prompt (safety filters) |
| Rate limit errors | Increase wait_seconds |
| No image in response | Check model name and API compatibility |
📝 License
MIT License
🙏 Credits
- Built for ComfyUI
- Powered by Google Gemini API