tencent/Hy-SimulMT-30B-A3B
tencent/Hy-SimulMT-30B-A3B
Tencent Mixun has released and open-sourced Hy-SimulMT-30B-A3B, supporting real-time Chinese-English two-way voice-to-voice simultaneous interpretation, outputting translation text and voice while receiving audio, and also supporting real-time voice cloning. In the RealSI long audio evaluation, its StreamLAAL is about 3 seconds, the lowest among the compared systems, and its GEMBA score is higher than the compared open-source models and four out of five commercial APIs.
The original text provides a comparison of the delay and quality of Chinese-English simultaneous interpretation, along with open-source weights and deployment methods, which can be used to judge the usability of real-time voice translation.
Source: Tencent Hunyuan · Hugging Face new models · huggingface.co