Voice AI in vogue: Nvidia releases new open transcription model

New open model for transcription

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.

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

Capabilities and benchmarks

Nvidia boasts that Nemotron 3.5 can translate 31 languages to and from English, handle recordings up to an hour long, and tell up to eight speakers apart. It claims to beat rivals on benchmarks, including Hugging Face’s Open ASR Leaderboard.

These tools target developers building apps for business meetings and medical transcription, two of AI’s most established footholds in the workplace.

Nvidia says developers can use the model, the updated Granary dataset, and tools like its NeMo Speech model to customize apps to recognize an industry’s jargon in each language.

Limitations and hardware

One drawback: Transcribe works with recorded audio, not livestreams.

Still, the company promises Transcribe, at 30 billion parameters, can run on a single chip, from the Jetson AGX Thor robotics controller and DGX Spark chip powering new laptops, to Blackwell datacenter chips.

It joins Nvidia’s open-weight voice models, including the super small yet powerful Parakeet (1.1 billion parameters).

Busy period for AI voice

Nvidia’s latest voice entry comes during a busy period of AI voice releases:

• Mistral shipped Voxtral 3 earlier this month
• ElevenLabs launched Realtime for instant speech in voice agents
• Salesforce released Launa-Omni, its voice model