Nvidia releases open Nemotron 3.5 Transcribe speech-to-text model
Chip and AI company Nvidia on Wednesday released Nemotron 3.5 Transcribe, a free open-weight model that transcribes audio in up to 40 languages. It also published version 2 of its open Granary speech dataset.
Nvidia, the US chipmaker and AI developer, on Wednesday, October 28, released a new speech recognition model called Nemotron 3.5 Transcribe. The free, open-weight model converts audio into text in up to 40 languages. It is aimed at developers building apps for business meetings and medical transcription, two of the most established uses of AI in the workplace.
Alongside the model, Nvidia released version 2 of Granary, its open-source speech dataset.
What the model can do
According to Nvidia, 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. The company claims it beats rivals on benchmarks, including Hugging Face's Open ASR Leaderboard, a ranking of automatic speech recognition systems.
Developers can use the model, the updated Granary dataset and tools such as NeMo Speech to customize apps to recognize an industry's jargon in each language.
Limits and hardware
One drawback is that Transcribe works with recorded audio rather than live streams. The model has 30 billion parameters, and Nvidia says it can run on a single chip, ranging from the Jetson AGX Thor robotics controller and the DGX Spark chip powering new laptops to Blackwell datacenter chips.
The model joins Nvidia's other open-weight voice models, including the small Parakeet, which has 1.1 billion parameters.
A busy field for voice AI
The release comes during a busy period for AI voice products. Mistral shipped Voxtral 3 earlier this month, ElevenLabs launched Realtime for instant speech in voice agents, and Salesforce released its voice model Launa-Omni.
