Providers

stt-1

The xAI speech-to-text model stt-1, with every setting it accepts, the allowed value and the default.

xAI's streaming recogniser, covering 25 languages. It ends an utterance on silence, so endpointing is the setting that shapes turn-taking.

Reached through the LiveKit Inference gateway.

One model, and the four settings below are everything it accepts.

Read this before using stt-1

xAI refuses any key it does not recognise. Where other providers ignore an unknown key, this one refuses the session, so the connection never opens and the call fails seconds after connecting. Only the four keys below may be present: a misspelled name, or a setting copied from another provider's transcriber, ends every call on that transcriber.

Settings#

Every setting is optional. Leave one out and OnCore sends nothing for it, so the Default applies. If invalid is what the provider does with a value it does not accept.

Setting	Type	Allowed	Default	If invalid
`endpointing`	int (ms)	0 to 5000	Provider's own	Refused, call fails
`format`	bool	true or false	Provider's own	Refused, call fails
`diarize`	bool	true or false	`false`	Used as sent
`interimResults`	bool	true or false	`true`	Used as sent

Overriding on a single call#

Return this from your assistant-request webhook to change these settings for one call only. The stored transcriber is left untouched, and the next call starts from it again.

json

{
  "data": {
    "overrides": {
      "transcriber": {
        "settings": {
          "endpointing": 300,
          "format": true,
          "diarize": false,
          "interimResults": true
        }
      }
    }
  }
}

Send only the keys you want to change: anything you leave out keeps the value the assistant already has. Add "id" beside "settings" to swap the whole transcriber for another one from your catalogue at the same time.

endpointing: Silence before an utterance is treated as final. Both bounds are enforced exactly: 0 and 5000 are accepted, -1 and 5001 are refused.

format: Inverse text normalisation, turning "one hundred dollars" into "$100". Setting it to true requires a language on the transcriber: without one the session is refused and the call fails. false is safe with or without a language.

These four are the complete set of settings this model accepts.

The language on the transcriber matters more here than on other providers. xai/stt-1 covers 25 languages, and format refuses to run without one being set.

Next steps#

Google: speech-to-text from Google.
Providers introduction: per-call overrides, and how a wrong value behaves across providers.

PreviousInk 2 Multilingual Nextgemini-3.5-transcribe-live