Skip to content
This repository was archived by the owner on Dec 26, 2025. It is now read-only.

README.md

Latest commit

 

History

History
90 lines (67 loc) · 2.85 KB

README.md

File metadata and controls

90 lines (67 loc) · 2.85 KB

Img2Img Example

English | 日本語

⚠️ Development Tool Notice

  • This is an internal development tool
  • May change frequently and contain bugs
  • Not officially supported
  • For production-level real-time research tools, use Livepeer Stream Model Lab

This example, based on this MPJEG server, runs image-to-image with a live webcam feed or screen capture on a web browser.

Features

  • Standard Mode: Basic image-to-image generation with SD-Turbo
  • ControlNet Mode: Enhanced generation with ControlNet support (depth, canny, pose, etc.)
  • Real-time streaming: WebRTC/WebSocket based streaming for low latency
  • Web interface: No desktop app required, runs in browser

Usage

You need Node.js 18+ and Python 3.10 to run this example. Please make sure you've installed all dependencies according to the installation instructions.

Standard Mode (Default)

cd frontend
npm i
npm run build
cd ..
pip install -r requirements.txt
python main.py --acceleration tensorrt

or

Quick Start Script

chmod +x start.sh
./start.sh

Then open http://0.0.0.0:7860 in your browser. (*If http://0.0.0:7860 does not work well, try http://localhost:7860)

ControlNet Configuration

When using ControlNet mode, you can specify:

  • Model: Base diffusion model (SD1.5, SD-Turbo, SDXL-Turbo)
  • ControlNets: One or multiple ControlNet models with preprocessors
  • Parameters: Generation settings, temporal consistency, acceleration options

See the ControlNet configuration examples for detailed YAML configuration options.

ControlNet Mode

To use ControlNet, provide a YAML configuration file:

python main.py --acceleration tensorrt --controlnet-config /path/to/config.yaml

Running with Docker

docker build -t img2img .
docker run -ti -e ENGINE_DIR=/data -e HF_HOME=/data -v ~/.cache/huggingface:/data  -p 7860:7860 --gpus all img2img

Where ENGINE_DIR and HF_HOME set a local cache directory, making it faster to restart the docker container.

Command Line Options

--host HOST                    Host address (default: 0.0.0.0)
--port PORT                    Port number (default: 7860)
--controlnet-config PATH       Path to ControlNet YAML configuration (optional)
--acceleration ACCEL           Acceleration type: none, xformers, sfast, tensorrt
--taesd / --no-taesd          Use Tiny Autoencoder (default: enabled)
--engine-dir DIR              TensorRT engine directory
--debug                       Enable debug mode