This project provides a framework for creating an AI-powered virtual YouTuber (VTuber) that can understand chat messages, respond intelligently, and display emotional expressions. It's primarily targeted at content creators looking to automate and enhance their live streaming presence, offering a simplified yet expressive AI avatar.

How It Works

The system comprises several microservices orchestrated via Docker Compose. Key components include a chat module for fetching live stream comments (Bilibili initially, with YouTube/Twitch planned), a ChatGPT-based chatbot for dialogue generation, an emotion analysis module for text sentiment, and a TTS engine for voice synthesis. A Live2D rendering engine displays the avatar, driven by a dedicated driver that translates chat interactions and emotional analysis into facial expressions and lip-syncing, using audio analysis or pre-defined motion playback.

Quick Start & Requirements

• Install: Clone the repository recursively (git clone --recursive) and use docker compose up -d.
• Prerequisites: Docker and Docker Compose.
• Configuration: Modify docker-compose.yml for proxy settings and configs/ directory for API keys (e.g., Azure TTS, ChatGPT) and room IDs.
• Resources: Requires ~1GB disk space for Docker images; network connectivity to Docker Hub.
• Docs: Configuration Details

Highlighted Details

• Fully Dockerized for cross-platform compatibility.
• Supports multiple Live2D lip-sync strategies: audio analysis, motion looping, or none.
• Offers API endpoints for manual control of avatar actions and dialogue.
• Configurable to use alternative ChatGPT-compatible APIs and HTTP proxies.

Maintenance & Community

• Project is actively developed with regular updates and CI/CD for Docker images.
• QQ group: 569686683.
• Open to contributions and discussions via GitHub Issues and PRs.

Licensing & Compatibility

• Licensed under MIT, permitting commercial use, modification, distribution, and private use without liability or warranty.

Limitations & Caveats

• The default Dockerfile is optimized for mainland China users; others should use gh.Dockerfile.
• Development environment setup is not Dockerized and requires specific Python/Node.js versions and package managers (pyenv, poetry, pnpm).
• TTS module currently has macOS dependencies, though other systems are supported via Docker.