NVIDIA opens multilingual transcription model that also diarizes speakers
What is NVIDIA Nemotron 3.5 Transcribe and Granary v2: Intelligent Multilingual Transcription You Can Prompt and Adapt?
The Nemotron 3.5 Transcribe model turns up to hour-long audio into text in 40 languages, adds speaker labels for up to eight voices, and can translate 31 languages to or from English. It is promptable, so developers can steer formatting, vocabulary and where the processing runs.
Who it is for
Anyone building meeting assistants, call-center analytics or medical dictation tools can use the free model to get accurate, multi-language transcripts with speaker IDs without training their own ASR.