NVIDIA Nemotron 3.5 Transcribe and Granary v2: Intelligent Multilingual Transcription You Can Prompt and Adapt

NVIDIA has released Nemotron 3.5 Transcribe, an open model that combines transcription, speaker attribution, translation and audio question answering. It transcribes 40 languages, translates 31 of them to or from English, labels up to eight speakers and handles recordings of about an hour. Output can be guided with natural-language prompts (format, domain vocabulary). Weights are on Hugging Face and ModelScope, with quantized versions for local use. NVIDIA also released the Granary v2 dataset and tools to adapt the model.

See original publication↗NVIDIA (Hugging Face officiel) · huggingface.co