Deepgram releases improved Nova-3 models for Swedish and Dutch speech-to-text
Nova-3 Swedish and Dutch Model Enhancements
Deepgram has released updated Nova-3 Swedish and Nova-3 Dutch models with improved accuracy for speech-to-text applications.
What's New
- Enhanced accuracy for both streaming and batch transcription modes
- Support for Swedish (sv, sv-SE) and Dutch (nl) language codes
- Models accessible via the existing Nova-3 model identifier
How to Use
To access the updated models, set the model parameter to nova-3 and specify the relevant language code:
- Swedish:
svorsv-SE - Dutch:
nl
No breaking changes—existing implementations will automatically benefit from the improved models. Developers can reference the Models and Language Overview documentation for additional details on supported languages and features.