Voice AI in vogue: Nvidia releases new open transcription model

Key points

• Nvidia released Nemotron 3.5 Transcribe, a free, open-weight model that transcribes audio in up to 40 languages.
• The model translates 31 languages to and from English, handles recordings up to an hour long and distinguishes up to eight speakers.
• Transcribe works with recorded audio rather than live streams and runs on a single chip at 30 billion parameters.

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

According to Nvidia, the model translates 31 languages to and from English, can process recordings up to an hour long and can tell up to eight speakers apart. Nvidia says it outperforms rivals on benchmarks, including Hugging Face's Open ASR Leaderboard.

Target users and limits

The tools are aimed at developers building apps for business meetings and medical transcription. Nvidia says developers can use the model, the updated Granary dataset and tools such as its NeMo Speech model to customize apps to recognize industry jargon in each language.

One drawback is that Transcribe works with recorded audio rather than live streams.

Hardware and crowded market

At 30 billion parameters, Nvidia says Transcribe 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 Parakeet at 1.1 billion parameters.

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 Launa-Omni.