NVIDIA Releases Nemotron 3.5 Transcribe, a Speech Model for 40 Languages
XORit Editorial · Sources, AI assistance and corrections
On October 28, 2026, NVIDIA introduced Nemotron 3.5 Transcribe, a model for transcribing and translating speech in 40 languages. It works with audio recordings up to one hour long.
The transcription style can be set with a plain-text instruction, without retraining the model. For example, users can ask it to preserve verbatim speech or write out numbers. The model also supports contextual transcription and transcription labeled by speaker.
On request, Nemotron 3.5 Transcribe adds timestamps for individual words and identifies speakers. NVIDIA released the model weights under the OpenMDW-1.1 license and a toolkit for fine-tuning in NeMo Speech.
Image: Kevin McCarthy CC BY 2.0 · License
Translation editor: Салтанов А.
Source: research.nvidia.com ↗