Cadence is a premium generative audio dashboard designed to showcase the Lit Audio UI components and state-of-the-art in voice-to-voice interaction and real-time audio orchestration. Built with a component-driven frontend and a high-concurrency Go backend, it serves as a "switchboard" for Google's most advanced speech and language APIs.
Key Features
- 3D Carousel Stage: A fluid, component-driven UI for switching between various audio engines and personas using lit-audio-ui components.
- Gemini Live Integration: Real-time, low-latency voice-to-voice interaction with a synchronized 3D Orb visualization.
- Streaming Transcription: Utilizes Speech v2 (Gemini TTS) for robust, real-time speech-to-text with persistent session history.
- Dual-Path Synthesis: Supports both Gemini Flash TTS for speed and Cloud TTS v2 Prompt API for character-driven, expressive performances.
- Switchboard Protocol: A custom JSON-over-WebSocket protocol that manages engine re-configuration without disconnecting the session.
The project demonstrates how to unify disparate audio streaming services into a single, cohesive user experience suitable for streaming audio AI agents.