ComfyUI IC-Custom Node

A custom node for ComfyUI that integrates IC-Custom model for high-quality image customization and generation.

✨ Features

• 🎨 High-Quality Image Generation: Powered by FLUX.1-Fill-dev and IC-Custom models
• 🖼️ Image Customization: Generate customized images based on reference images
• 🎯 Flexible Generation Modes: Support for position-free and position-precise generation
• ⚙️ Advanced Controls: Configurable guidance scale, inference steps, and seed control
• 🚀 Optimized Performance: Model quantization and offloading for better memory efficiency

📦 Installation

Step 1: Install the Node

# Navigate to ComfyUI custom_nodes directory
cd ComfyUI/custom_nodes

# Clone the repository
git clone https://github.com/HM-RunningHub/ComfyUI_RH_ICCustom

# Install dependencies
cd ComfyUI_RH_ICCustom
pip install -r requirements.txt

Step 2: Download Required Models

Create the following directory structure in your ComfyUI models folder:

Main Models

FLUX.1-Fill-dev Model:

• Download: FLUX.1-Fill-dev
• Files: ae.safetensors, flux1-fill-dev.safetensors
• Location: ComfyUI/models/black-forest-labs/FLUX.1-Fill-dev/

IC-Custom Model:

• Download: IC-Custom
• Files: All files from the repository
• Location: ComfyUI/models/IC-Custom/

FLUX Redux Model:

• Download: FLUX.1-Redux-dev
• File: flux1-redux-dev.safetensors
• Location: ComfyUI/models/IC-Custom/

CLIP Models

SigLIP Model:

• Download: siglip-so400m-patch14-384
• Files: All files from the repository
• Location: ComfyUI/models/clip/siglip-so400m-patch14-384/

XFlux Text Encoders:

• Download: xflux_text_encoders
• Files: All files from the repository
• Location: ComfyUI/models/clip/xflux_text_encoders/

CLIP Vision Models

CLIP ViT Large:

• Download: clip-vit-large-patch14
• Files: All files from the repository
• Location: ComfyUI/models/clip_vision/clip-vit-large-patch14/

SigCLIP Vision:

• Download: sigclip_vision_patch14_384
• File: sigclip_vision_patch14_384.safetensors
• Location: ComfyUI/models/clip_vision/

🚀 Usage

Basic Workflow

1. Add Model Loader: Add "RunningHub ICCustom Loader" node to your workflow
2. Add Sampler: Add "RunningHub ICCustom Sampler" node and connect the pipeline output
3. Configure Inputs:
   • Connect reference image
   • Set prompt text
   • Configure generation parameters
   • Optionally add target image and mask for precise control

Example Workflow

[Reference Image] → [ICCustom Loader] → [ICCustom Sampler] → [Save Image]
                                    ↓
                               [Prompt Input]

Generation Modes

• Position-Free: Generate without target constraints (no mask required)
• Position-Precise: Generate with specific target positioning (requires mask)

⚙️ Parameters

• Prompt: Text description for the generated content
• Guidance: Controls adherence to prompt (default: 40.0)
• True GS: Additional guidance parameter (default: 3.0)
• Steps: Number of inference steps (default: 25)
• Seed: Random seed for reproducible results

🔧 Requirements

• GPU Memory: 16GB+ VRAM recommended
• System RAM: 32GB+ recommended
• Storage: ~100B for all models
• Dependencies: PyTorch, Diffusers, Transformers

📄 License

This project is licensed under the Apache 2.0 License.

🔗 References

• IC-Custom
• FLUX Models
• ComfyUI

🙏 Acknowledgments

Special thanks to AIwood爱屋研究室 (Bilibili) for helping with Windows environment testing and contributing to the installation documentation. <img width="1464" height="1140" alt="image" src="https://github.com/user-attachments/assets/f77bf3e1-ece3-4c70-9bbe-2e77044cfa1e" /> <img width="1706" height="1119" alt="image" src="https://github.com/user-attachments/assets/71cd94aa-7f45-4fc3-a06d-37105cf7daae" />

🤝 Contributing

Contributions are welcome! Please feel free to submit issues and pull requests.