The improved Nova-3 Portuguese model is here!

[naomi-deepgram]

Portuguese transcription is now more accurate across batch and streaming speech-to-text, including both Brazilian Portuguese and European Portuguese.

Latest improvements include:

• 19% lower overall WER for batch
• 16% lower overall WER for streaming

Available now with: model="nova-3"
Supported language codes: pt, pt-BR, pt-PT

Try it in the Playground or get started in the docs:
https://playground.deepgram.com/?endpoint=listen-streaming&endpointing=10&interim_results=true&smart_format=true&language=pt&model=nova-3