Voice AI in vogue: Nvidia releases new open transcription model
On this page
Nvidia released Nemotron 3.5 Transcribe, a free, open-weight model that transcribes audio in up to 40 languages and translates 31 languages to and from English. The company also released version 2 of Granary, its open source speech dataset. Nvidia says the 30-billion-parameter model can handle recordings up to an hour long, separate up to eight speakers, and run on a single chip from its Jetson AGX Thor robotics controller and DGX Spark laptops up to Blackwell datacenter chips. It claims to beat rivals on benchmarks including Hugging Face’s Open ASR Leaderboard. The release targets developers building business-meeting and medical-transcription apps, two of AI’s most established workplace footholds. The model works with recorded audio rather than livestreams. It joins Nvidia’s open-weight voice models such as Parakeet, and comes amid a burst of voice releases from Mistral’s Voxtral 3, ElevenLabs’ Realtime, and Salesforce’s Launa-Omni.
Why It Matters
Nvidia’s open transcription model is a frontier-lab release in the fast-moving voice-AI space, and it signals how frontier labs are pushing open-source and open-weight models into speech. The combination of a permissive model, an open dataset, and single-chip deployment lowers the cost of building transcription and translation into workplace tools, an important step for multimodal and generative capabilities. It also illustrates the widening competition among labs racing to ship voice models for agents and autonomy and for enterprise applications.
Sources
- TechCrunch: https://techcrunch.com/2026/10/28/voice-ai-in-vogue-nvidia-releases-new-open-transcription-model/
Back to Recent News · Sub-article index.