NVIDIA releases Nemotron 3.5 Transcribe and Granary v2 for multilingual transcription
NVIDIA published a Hugging Face blog post introducing Nemotron 3.5 Transcribe and Granary v2, multilingual transcription models that can be prompted and adapted. The announcement is based on the post's title and metadata only; specific benchmarks, languages, and licensing details were not available in the provided source.
Why it matters
Builders producing transcripts, subtitles, or searchable audio content for multilingual audiences may have a new open transcription option to evaluate against current speech-to-text tools.
Who should care
Ecommerce brands with video or podcast content, media producers, customer support teams handling voice calls, developers building speech-to-text features
What you can do
Open the NVIDIA Hugging Face post, check the listed languages, license, and benchmark results, and test the model on a sample of your own multilingual audio.