Key Info

Tencent HY has open-sourced Hy-SimulMT, a real-time Chinese-to-English and English-to-Chinese simultaneous interpreting model, available on Hugging Face, ModelScope, and GitHub.

Highlights

  • Low latency of roughly 3 seconds for simultaneous interpreting
  • Real-time voice cloning capability
  • The company claims it outperforms most open-source models and commercial APIs
  • Weights and code released via Hugging Face, ModelScope, and GitHub