Nvidia releases open-weight Nemotron 3.5 model for multilingual audio transcription
Nvidia has debuted Nemotron 3.5 Transcribe, a free open-weight speech-to-text model supporting up to 40 languages on recorded audio alongside the open-source Granary v2 dataset.
Nvidia has launched Nemotron 3.5 Transcribe—also referred to as Nemotron 3.5 ASR—a free, open-weight artificial intelligence model that can transcribe audio across up to 40 languages. Alongside the model release, the company introduced Granary v2, an open-source speech dataset.[1][2][3][6]
The release features approximately 4.3 billion active parameters and supports both multilingual audio transcription and translation. According to Nvidia, Nemotron 3.5 Transcribe ranks first among open-weight models on Hugging Face's English transcription leaderboard and can identify individual speakers even when they talk over one another.[1][2][5]
The model operates only on pre-recorded audio and does not support live streams. TechCrunch reported that this limitation directs the technology toward meeting and medical transcription workflows while leaving real-time audio tasks to competing platforms.[4]
Key facts
- Nvidia released Nemotron 3.5 Transcribe (also designated Nemotron 3.5 ASR) as a free, open-weight speech model.
- The model supports transcription across up to 40 languages as well as audio translation.
- Nemotron 3.5 Transcribe contains approximately 4.3 billion active parameters.
- Nvidia reported that the model ranks first among open-weight models on Hugging Face's English transcription leaderboard.
- The model can distinguish which speaker is talking, including when multiple people speak at the same time.
- The tool processes only recorded audio files and does not function on live audio feeds.
- Nvidia launched an open-source speech dataset called Granary v2 alongside the model.
Sources · 6 sources
- TE
TechmemeArticle ·
Nvidia debuts Nemotron 3.5 Transcribe, an open-weight model for multilingual audio transcription and translation, and Granary v2, its open-source speech dataset (Kyt Dotson/SiliconANGLE) Kyt Dotson / SiliconANGLE : Nvidia debuts Nemotron 3.5 Transcribe, an open-weight model for multilingual audio transcription and translation, and Granary v2, its open-source speech dataset — Nvidia Corp. today released a leaderboard-winning open artificial intelligence model for multilingual audio transcription and translation …
Open source - WS
Wall St Engine@wallstenginePost on X ·
NVIDIA JUST RELEASED NEMOTRON 3.5 TRANSCRIBE 40 languages. ~4.3B active parameters. Open weights. $NVDA reports #1 among open-weight models on Hugging Face’s English transcription leaderboard. Watch it track who said what, even when speakers talk over each other. https://t.co/JU68mPUwR3
Open source - CB
Crypto BriefingArticle ·
Nvidia releases Nemotron 3.5 ASR, a free open-weight model for 40-language audio transcription Nvidia's Nemotron 3.5 ASR democratizes multilingual audio transcription, potentially transforming global communication and accessibility. The post Nvidia releases Nemotron 3.5 ASR, a free open-weight model for 40-language audio transcription appeared first on Crypto Briefing .
Open source - TE
TechCrunch@TechCrunchPost on X ·
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 https://t.co/58l5TUPMsl
Open source - SI
SiliconANGLE@SiliconANGLEPost on X ·
Nvidia releases all-in-one adaptable Nemotron 3.5 audio transcription and translation model https://t.co/3gfzJP1lcn
Open source - TE
Techmeme@TechmemePost on X ·
Nvidia debuts Nemotron 3.5 Transcribe, an open-weight model for multilingual audio transcription and translation, and Granary v2, its open-source speech dataset (@kytsune / SiliconANGLE) (Visit Techmeme dot com for the link and full context!)
Open source