Nvidia has introduced a new open-source model called Nemotron 3.5 Transcribe. The system is offered free of charge and includes publicly available weights. It can directly translate audio in thirty‑one languages and also from English into those languages. The model is capable of distinguishing up to eight different speakers within a recording. It can handle audio files that are as long as one hour. Nvidia promotes the tool as supporting transcription in forty languages overall.