What was announced
NVIDIA announced Nemotron 3.5 Transcribe, a speech recognition model, and Granary v2 on the Hugging Face blog. According to the headline, the model can be prompted and adapted for multilingual transcription.
Why it matters
Speech recognition underpins many applications, from voice assistants to meeting summaries. A multilingual, promptable model could make it easier for developers to cover different languages and use cases with a single model.
Limits of the information
This report is based only on the announcement headline; details such as parameter count, supported languages, license and access terms are not in the source. Performance claims are therefore omitted.
