StreamDiffusion Node for ComfyUI

A real-time StreamDiffusion pipeline node for ComfyUI with integrated ControlNet, IPAdapter, and LoRA support.

Features

• Real-time, multi-control Stable Diffusion workflows
• Integrated ControlNet, IPAdapter, and LoRA support
• TensorRT acceleration for high performance
• Flexible configuration and batching
• Dynamically change parameter values at runtime via the streaming sampler node

Installation

Using ComfyUI Registry (Recommended)

comfy node install streamdiffusion

Manual Installation

Clone this repository into your ComfyUI custom nodes directory:

git clone https://github.com/RUFFY-369/ComfyUI-StreamDiffusion.git

Install dependencies:

pip install -r requirements.txt

Usage

After installation, launch ComfyUI. The StreamDiffusion node and its features will be available in the node menu.

Example Workflows

1. Build TensorRT Engines

   • Use the workflow: examples/sd15_tensorrt_engine_build.json
   • This workflow demonstrates how to build TensorRT engines for your models within ComfyUI.
   • Alternatively, you can run the script scripts/build_tensorrt_engines.py directly to build engines without using the ComfyUI workflow.

2. Inference with Streaming Sampler

   • Use the workflow: examples/sd15_all_dynamic_params_wlora.json
   • This workflow shows how to use the built engines for real-time inference and dynamically change parameters with the streaming sampler node.

To use these workflows:

1. Open ComfyUI.
2. Load the desired workflow JSON file from the examples/ directory.
3. Follow the node connections and adjust parameters as needed.

Requirements

• Python 3.10+
• NVIDIA GPU (for TensorRT acceleration)
• ComfyUI >= 1.0.0

Support & Documentation

• Documentation (README)
• Bug Tracker

License

This project is licensed under the terms of the Apache License 2.0, the same as the original StreamDiffusion repository.

Citation

If you use this node or its ideas in your research or projects, please cite the original StreamDiffusion paper:

StreamDiffusion: Real-Time Text-to-Image Generation on Video and Web Cameras (arXiv:2312.12491)

@article{zhang2023streamdiffusion,
	title={StreamDiffusion: Real-Time Text-to-Image Generation on Video and Web Cameras},
	author={Zhang, Yifan and others},
	journal={arXiv preprint arXiv:2312.12491},
	year={2023}
}