LiveTalking: Build Commercial-Grade AI Digital Humans with Open Source

Transform Text into Lifelike Talking Avatars

LiveTalking (7.2k ⭐️) delivers production-ready real-time digital humans that sync audio, lip movements, and facial expressions with commercial quality. Originally metahuman-stream, this Python powerhouse supports multiple cutting-edge models and WebRTC streaming for seamless browser integration.

🚀 Core Features

• 4+ Digital Human Models: wav2lip (60 FPS on RTX 3060), musetalk (72 FPS on 4090), ernerf, Ultralight-Digital-Human
• Voice Cloning: Real-time TTS with interruption support
• WebRTC + Virtual Camera: Browser-compatible streaming
• Multi-concurrency: Scale across CPU/GPU resources
• Custom Avatars: Upload your own character images

🎯 Quick Start (5 Minutes)

# Ubuntu 24.04 + Python 3.10 + CUDA 12.4
conda create -n livetalking python=3.10
conda activate livetalking
conda install pytorch==2.5.0 torchvision==0.20.0 pytorch-cuda=12.4 -c pytorch -c nvidia
pip install -r requirements.txt

# Download models (Quark/Google Drive links)
python app.py --transport webrtc --model wav2lip --avatar_id wav2lip256_avatar1

Browser test: http://your-server:8010/webrtcapi.html → Type → Watch AI speak!

🐳 Docker (Zero Setup)

docker run --gpus all -it --network=host registry.cn-beijing.aliyuncs.com/codewithgpu2/lipku-metahuman-stream:2K9qaMBu8v

⚡ Performance Benchmarks

💎 Commercial Extensions Available

• HD wav2lip models
• Real-time subtitles + interruption
• Multi-avatar per stream
• Camera-driven expressions
• Unlimited avatar duration

🎮 Use Cases

1. Live Streaming: Interactive AI co-hosts
2. Education: Multilingual tutors
3. Customer Service: 24/7 AI agents
4. Content Creation: Automated talking heads
5. Virtual Events: Scalable digital presenters

📦 One-Click Cloud Deployment

• UCloud/AutoDL mirrors available
• Pre-configured GPU instances
• Enterprise documentation: livetalking-doc.readthedocs.io

Get started: GitHub - lipku/LiveTalking ⭐️ + 🚀 = Commercial AI avatars in minutes!

Apache 2.0 licensed • 1.1k forks • Active community