The 30-billion-parameter model handles 40 languages and recorded audio, with an updated Granary speech dataset for developers.
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. The 30-billion-parameter model translates 31 languages to and from English, handles recordings up to an hour long, and distinguishes up to eight speakers.
The model joins Nvidia’s open-weight voice lineup, which includes the smaller Parakeet model at 1.1 billion parameters. Nvidia claims Nemotron 3.5 Transcribe beats rivals on benchmarks, including Hugging Face’s Open ASR Leaderboard, and the release lands during a busy stretch of voice AI launches from Mistral, ElevenLabs, and Salesforce.
Developers can pair the model, the updated Granary dataset, and NeMo Speech tools to tune apps for industry jargon in each supported language. Business meetings and medical transcription give teams two established workplace uses, while recorded-audio limits keep live-streaming agents out of scope for now.
Nvidia says the 30-billion-parameter model can run on a single chip, spanning the Jetson AGX Thor robotics controller, the DGX Spark chip for new laptops, and Blackwell datacenter chips. Watch for live audio support and for how quickly developers adopt the model against recent voice releases from Mistral, ElevenLabs, and Salesforce.
What matters
- Nvidia released Nemotron 3.5 Transcribe and version 2 of its open-source Granary speech dataset.
- Builders can customize meeting and medical transcription apps with the model, dataset, and NeMo Speech tools.
- Watch whether Nvidia adds live-stream support and how rivals like Mistral and ElevenLabs respond.
Why it matters
Watch whether Nvidia adds live-stream support and how rivals like Mistral and ElevenLabs respond.
This GenAI News article was prepared in original wording using reporting and materials published by TechCrunch AI. Source reference: https://techcrunch.com/2026/10/28/voice-ai-in-vogue-nvidia-releases-new-open-transcription-model/.
Drafted by the GenAI News review pipeline.