Skip to content
Advertisement
Nvidia releases Nemotron 3.5 Transcribe, a free open-weight model for 40-language transcription

Nvidia releases Nemotron 3.5 Transcribe, a free open-weight model for 40-language transcription

The model handles hour-long recordings and up to eight speakers, but works only on recorded audio, not live streams.

Nvidia released Nemotron 3.5 Transcribe, a free, open-weight model that transcribes audio in up to 40 languages, TechCrunch reported. The company also released version 2 of Granary, its open-source speech dataset.

Advertisement

The model can translate 31 languages to and from English, handle recordings up to an hour long and distinguish up to eight speakers. Nvidia says it beats rivals on benchmarks including Hugging Face’s Open ASR Leaderboard. It has 30 billion parameters and can run on a single chip, from Jetson AGX Thor to Blackwell data center GPUs.

Built for recordings, not live audio

The model works only on recorded audio, not live streams, leaving real-time use to rivals such as ElevenLabs, TechCrunch said. Nvidia is targeting developers building business meeting and medical transcription apps, who can use Granary and its NeMo Speech tools to tailor the model to industry jargon.

Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.