NVIDIA Introduces Promptable Multilingual Transcription Models
NVIDIA has introduced two models for multilingual transcription: Nemotron 3.5 Transcribe and Granary v2. According to the announcement, both are designed to support promptable and adaptable transcription workflows.
The release points to a broader effort to make speech-recognition systems more flexible across languages and tasks. By allowing users to guide or adapt model behavior, NVIDIA is positioning the models as tools for a wider range of transcription applications rather than fixed, task-specific systems.