Models · Updated 28 Oct, 09:19 pm IST
Nvidia releases Nemotron 3.5 Transcribe speech model
Why it matters for readers: It demonstrates a single, efficient model that combines several speech tasks—transcription, translation and speaker ID—so you can test it on real audio.
- Nemotron 3.5 Transcribe handles transcription, translation and speaker identification in one model.1
- It transcribes 40 languages, translates 31 languages to or from English, and identifies speakers for 13 languages.1
- The model has 30 billion parameters but activates about 4 billion per step, letting it run on a single Nvidia GPU.1
- Recordings up to an hour long with up to eight speakers are supported, and users can control outputs with simple instructions and hints for technical terms.1
- Nvidia released a large training dataset and tools to adapt the model for specialized domains.1
Get a brief like this every morning
Uzha reads hundreds of sources and gives you the stories that matter for your work, with every source linked. Free.
Get started