ComfyUI OmniSVG Nodes

Generate high-quality SVG graphics from text descriptions and images using OmniSVG in ComfyUI.

🎯 Features

• 🔧 OmniSVG Model Loader: Load and manage OmniSVG models
• 📝 Text to SVG: Generate SVG graphics from text descriptions
• 🖼️ Image to SVG: Convert images to SVG format
• 🔄 SVG to Image: Convert SVG strings to ComfyUI IMAGE tensors
• 💾 SVG Saver: Save SVG files to disk

🚀 Quick Start

# Clone to ComfyUI custom_nodes directory
cd ComfyUI/custom_nodes/
git clone https://github.com/yourusername/ComfyUI-OmniSVG.git
cd ComfyUI-OmniSVG

# Install dependencies and setup
python install.py

# Restart ComfyUI

📋 Table of Contents

• Installation
• Requirements
• Usage
• Node Reference
• Examples
• Troubleshooting
• Contributing
• License

📦 Installation

Method 1: Automated Installation (Recommended)

# Navigate to ComfyUI custom_nodes directory
cd ComfyUI/custom_nodes/

# Clone the repository
git clone https://github.com/yourusername/ComfyUI-OmniSVG.git
cd ComfyUI-OmniSVG

# Run automated installer
python install.py

# Restart ComfyUI

The installer will:

• ✅ Install all Python dependencies
• ✅ Set up the models directory
• ✅ Verify the installation
• ✅ Provide next steps

Method 2: Manual Installation

<details> <summary>Click to expand manual installation steps</summary>

1. Clone Repository

   cd ComfyUI/custom_nodes/
   git clone https://github.com/yourusername/ComfyUI-OmniSVG.git
   cd ComfyUI-OmniSVG
   

2. Install Dependencies

   pip install -r requirements.txt
   

3. Download Model Weights

   # Create models directory
   mkdir -p ../../models/omnisvg/
   
   # Download model (requires huggingface-cli)
   huggingface-cli download OmniSVG/OmniSVG --local-dir ../../models/omnisvg/OmniSVG-3B
   

4. Restart ComfyUI

</details>

Method 3: Using Existing OmniSVG Installation

If you already have OmniSVG installed (e.g., in /mnt/storage/OmniSVG-3B), the nodes will automatically detect and use it.

Requirements

• GPU Memory: 17GB+ VRAM for optimal performance
• Python: 3.8+
• PyTorch: 2.3.0+
• CUDA: 12.1+ (recommended)

Usage

Basic Workflow

1. Load Model: Use "OmniSVG Model Loader" to load the model
2. Generate SVG:
   • Use "OmniSVG Text to SVG" for text-based generation
   • Use "OmniSVG Image to SVG" for image conversion
3. Process Output:
   • Use "SVG to Image" to convert SVG to ComfyUI images
   • Use "SVG Saver" to save SVG files

Node Parameters

OmniSVG Text to SVG

• text_prompt: Description of the SVG to generate
• temperature: Controls randomness (0.1-2.0)
• top_p: Nucleus sampling parameter (0.1-1.0)
• top_k: Top-k sampling parameter (1-100)
• repetition_penalty: Prevents repetition (1.0-2.0)

OmniSVG Image to SVG

• image: Input image to convert
• target_size: Resize image before processing (64-512px)
• temperature: Lower values for more faithful conversion (0.01-1.0)
• top_p: Nucleus sampling (0.001-1.0)

Model Storage

Models should be placed in:

ComfyUI/models/omnisvg/
└── OmniSVG-3B/
    ├── config.yaml
    ├── config.json
    ├── pytorch_model.bin
    └── README.md

Performance Tips

• GPU: Use CUDA for 10-50x faster generation
• Memory: Ensure 17GB+ VRAM available
• Batch Size: Process one image at a time for stability
• Temperature: Lower values (0.1-0.3) for more consistent results

Troubleshooting

Common Issues

1. "Download Required" in model list

   • Download OmniSVG models to ComfyUI/models/omnisvg/
   • Or ensure existing installation is accessible

2. CUDA Out of Memory

   • Free up GPU memory from other processes
   • Use CPU mode (slower but functional)

3. Import Errors

   • Install all requirements: pip install -r requirements.txt
   • Ensure OmniSVG core files are copied correctly

Error Messages

• "Model not found": Check model path and file permissions
• "Generation error": Check GPU memory and model loading
• "SVG conversion failed": Verify CairoSVG installation

Examples

Text to SVG

Input: "A minimalist icon of a coffee cup with steam"
Output: Clean SVG icon suitable for web use

Image to SVG

Input: Photo of a simple logo or icon
Output: Vector SVG representation

License

This project follows the same license as OmniSVG. Please refer to the original OmniSVG repository for licensing details.

Credits

• OmniSVG: Original model and implementation
• ComfyUI: Node framework and integration
• Qwen2.5-VL: Base vision-language model

Support

For issues and questions:

1. Check this README and troubleshooting section
2. Review ComfyUI console output for error details
3. Open an issue on GitHub with error logs