Voice AI in vogue: Nvidia releases new open transcription model
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 unveiled Nemotron 3.5 Transcribe, a free, open-weight transcription model capable of processing audio in up to 40 languages, on Wednesday. The company also released version 2 of Granary, its open-source speech dataset. According to Nvidia, Nemotron 3.5 can translate 31 languages to and from English, manage recordings up to an hour in length, and distinguish up to eight speakers simultaneously. The model claims to outperform competitors on benchmarks, including Hugging Face’s Open ASR Leaderboard.
Nvidia aims to cater to developers creating applications for business meetings and medical transcription, two prominent areas where AI has established a strong presence in the workplace. The company encourages developers to utilize Nemotron 3.5, the updated Granary dataset, and tools like its NeMo Speech model to fine-tune their applications to recognize industry-specific terminology in each language.
However, Transcribe is limited to processing recorded audio, not live streams. Despite this limitation, Nvidia asserts that Transcribe, with its 30 billion parameters, can run on a single chip, ranging from the Jetson AGX Thor robotics controller and DGX Spark chip powering new laptops to the Blackwell datacenter chips. This new voice release by Nvidia arrives amidst a flurry of AI voice model launches, such as Mistral's Voxtral 3, ElevenLabs' Realtime for instant speech generation in voice agents, and Salesforce's Launa-Omni, its voice model.
Written by urgent.news from TechCrunch's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
