TaDiCodec-TTS-AR-Qwen2.5-0.5B Demo

This is a demo of the TaDiCodec Text-to-Speech model with Qwen2.5-0.5B backbone.

Features:

  • Voice cloning with reference audio
  • Code-switching support (e.g., mixing English and Chinese)
  • Extremely low bitrate (0.0875 kbps)
  • High-quality speech generation

Example Inputs

Examples
Text to Synthesize Reference Text Reference Audio