🎭 Linly-Gemini Live: Real-Time AI Avatar(PersonasAI)

---

✨ Features

• 🗣️ Real-Time Interaction: Engage in seamless voice conversations with ultra-low latency powered by Gemini Live.
• 👄 Lip-Sync Animation: Uses MuseTalk to generate realistic lip movements synced perfectly to the AI's audio response.
• 👤 Custom Avatars: Upload any portrait image or video to create your own unique persona instantly.
• 🔄 Self-Healing Deployment: Automatically detects and downloads missing model weights at runtime, ensuring robust deployments on Hugging Face Spaces.
• 🛠️ Zero-Config Interface: Simple 2-button UI meant for immediate interaction—just "Start Streaming" and go.

---

🏗️ Architecture

This project combines state-of-the-art storage, inference, and LLM technologies:

1. Frontend (Gradio): Captures microphone input and renders the video stream.
2. Orchestrator (Python): manages the WebSocket connection and audio buffering.
3. Brain (Gemini Live): Processes audio input and generates intelligent, conversational audio responses via a dedicated WebSocket bridge (hosted on Railway).
4. Face Engine (MuseTalk): Takes the audio from Gemini and the video frames of the avatar to generate lip-synced video in real-time.