Tencent announces Hy-SimulMT for live interpretation

A formal announcement after earlier uploads

Tencent’s release notice dates the formal launch to October 26. The repository history records an October 24 weight upload, and that revision contains the weight shards and license. This event records the dated announcement, rather than claiming that downloads first became public on October 26; commit timestamps alone do not establish historical repository visibility. [1][2][3][4]

The project processes incoming audio and streams translated text between Chinese and English. Its inference stack requires a separately configured text-to-speech service for spoken output and voice cloning. The weights use Apache 2.0; Tencent’s approximately three-second StreamLAAL result measures translation lag in its evaluation, not time to first spoken output. [1][5]