Deepgram

Route to Deepgram by prefixing the model with deepgram/. Aura is a low-latency streaming model.

Models

Requests longer than the model's input limit are chunked server-side — see Synthesise speech.

Usage

curl -X POST https://api.speechbase.ai/v1/audio/speech \
  -H "Authorization: Bearer $SPEECHBASE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "mode": "inline",
    "model": "deepgram/aura-2",
    "voice": "aura-asteria-en",
    "text": "Hello from Speechbase!",
    "output": "mp3"
  }' --output hello.mp3

voice is a Deepgram voice ID of the form aura-<name>-en, for example aura-asteria-en, aura-luna-en, or aura-orion-en.

Timestamps

Deepgram has no native word alignment, so the /v1/audio/speech/with-timestamps endpoint routes the audio through the gateway's STT fallback to produce timestamps.

Provider options

Anything in providerOptions is forwarded to the Deepgram speak API unchanged.