Speech to Text
A speech-to-text provider powers the microphone button in the chat, so users can dictate a message instead of typing it. The Diarization badge marks providers that can also tell speakers apart, which matters when you transcribe meetings.
Available providers
On Premise — Self-hosted transcription service — no audio leaves your infrastructure, and speaker diarization is supported.
OpenAI — Transcribes with OpenAI Whisper through the OpenAI API. platform.openai.com
Mistral — Transcribes with Mistral Voxtral through the Mistral API. EU-hosted. console.mistral.ai
Note
Only the On Premise provider can tell speakers apart. With OpenAI or Mistral configured, audio is still transcribed, but asking for speaker diarization is refused.
Info
Until a default speech-to-text provider is configured for a scope, the microphone is hidden from the chat composer and the Audio connector is not offered when adding a data source. Configuring the first provider switches the microphone back on if it had been turned off.