Voice AI in vogue: Nvidia releases new open transcription model

Nvidia has released Nemotron 3.5 Transcribe, a free, open-weight model capable of transcribing audio in up to 40 languages and translating 31 languages to and from English. It can process recordings up to an hour long and distinguish up to eight speakers. Nvidia also updated its Granary open-source speech dataset to version 2. The 30 billion parameter model is designed for developers building applications for business meetings and medical transcription, and can run on various Nvidia chips, from Jetson AGX Thor to Blackwell datacenter chips.

Unlike other recent voice AI releases, Nvidia's Nemotron 3.5 Transcribe is an open-weight model that claims to outperform rivals on benchmarks like Hugging Face's Open ASR Leaderboard.