Nvidia releases open-weight Nemotron 3.5 Transcribe speech model
Original title: Voice AI in vogue: Nvidia releases new open transcription model
Nvidia also shipped Granary version 2 and says the model handles recorded audio only, not live streams.
Nvidia released Nemotron 3.5 Transcribe, a free open-weight model that transcribes audio in up to 40 languages, and version 2 of its open-source Granary speech dataset. Nvidia says the model can translate 31 languages to and from English, handle recordings up to an hour, separate up to eight speakers, and beat rivals including Hugging Face’s Open ASR Leaderboard. It targets meeting and medical transcription apps and, at 30 billion parameters, is said to run on one chip from Jetson AGX Thor and DGX Spark to Blackwell datacenter chips.
Unconfirmed: Benchmark wins, language coverage and the 30-billion-parameter size are Nvidia’s claims as reported by TechCrunch.
- Transcribe does not support live audio streams.
- Developers can customize jargon using the model, Granary and NeMo Speech tools.
- It joins Nvidia voice models including Parakeet, listed at 1.1 billion parameters.
- Recent voice releases cited include Mistral Voxtral 3, ElevenLabs Realtime and Salesforce Launa-Omni.
- Vendors
- Nvidia
- Organizations
- MistralElevenLabsSalesforce
From the original article
1 min read at techcrunch.com
Nvidia just released a free speech-to-text model that can transcribe up to 40 languages, but it only works on recordings, not live audio, so it’s a big step for meeting and medical transcription tools while still leaving real-time use to rivals like ElevenLabs.
By Julie Bort · techcrunch.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.