Skip to main content
Speech-to-text (STT) models listen to spoken audio and turn it into written text, as it happens. In a SLNG voice agent this is the “Listen” step: the caller speaks, the STT model transcribes it, and the language model reads that text to decide how to respond. How fast and accurate the STT model is affects everything after it. If it mishears the caller, the agent answers the wrong thing, and if it’s slow, the whole conversation feels delayed. STT providers we support:
  • Fish Audio
  • Deepgram
  • Soniox
  • Sarvam AI
  • Speechmatics
  • Reson8
  • Gradium
Get started quick with this guide.