Updated Nova-2 Model (English, Pre-recorded) - Improved Word Recognition
model = nova-2
Deepgram has done additional training to our Nova-2 General Model for English to improve accuracy for pre-recorded transcription, specifically addressing an issue with missing words / phrases. We intend to roll out similar updates for streaming transcription and across other Nova-2 tier models over the next two weeks.
There should be no code changes required on your end - all requests with model=nova-2 will automatically get the latest version with the fix applied.
For On-Premises customers, please reach out to your Deepgram contact if you'd like to update your models.