Skip to main content
​
April 9, 2026
FeatureUpdate
API v0.20.0

​
Language selection for SLNG-hosted Nova 3 STT

SLNG-hosted Deepgram Nova 3 speech-to-text endpoints now accept a language parameter in the WebSocket init config. Each language variant supports its own set of locale codes:
  • English — en, en-au, en-us, en-nz, en-gb, en-in
  • Spanish — es, es-us, es-419, es-ar, es-mx, es-es
  • Hindi — hi, en (English is now also supported on the Hindi variant)
  • Multi-language — multi
This gives you explicit control over the recognition language and locale, improving accuracy for regional accents and dialects. See the Speech-to-Text API reference for the full list of supported models and parameters.

​
English support on Nova 3 Hindi

The SLNG-hosted Deepgram Nova 3 Hindi endpoint now also accepts language: "en" for English transcription. This means you can use a single Hindi-region deployment for both Hindi and English audio without switching endpoints.

​
Expanded sample rates for Rime Arcana TTS

Rime Arcana text-to-speech now supports a wider range of audio sample rates: 8 kHz, 16 kHz, 22.05 kHz, 24 kHz (default), 32 kHz, 44.1 kHz, and 48 kHz. This makes it easier to match your application’s audio pipeline without resampling. See the Rime Arcana voice page for details.

​
Endpointing parameter simplified

The endpointing parameter on Deepgram STT endpoints has been simplified. It now accepts only an integer value (milliseconds of silence before finalizing speech). To disable endpointing, set it to 0 instead of false. The default remains 10.