Send text instead of audio

When your platform already runs speech recognition, send the words and skip the audio. This removes the transcription cost from the call and works on any platform that can open a WebSocket.

How

Open the socket and send the start message as on Connect your switch.

Send one text frame per finished utterance:

{"type": "transcript", "text": "This is the IRS. There is a warrant for your arrest.", "track": "inbound"}

Field	Default	Meaning
`text`	required	The words. Empty text is ignored.
`track`	`inbound`	`inbound` is the remote caller. `outbound` is the local party, labelled `callee`.

The server treats every transcript message as one final utterance. Do not send interim results. Extra fields are ignored.

Rules

You can send audio and text on the same socket. Both feed one transcript.

Send the caller's words. Send the called party's words as outbound only when you want them in the transcript. They are not scored.

Events, verdicts, and session.ended are the same as for audio. See Events and webhooks.

The filter is still charged per started minute of session time.

After the call

To score a finished transcript in one request, use Score a transcript or a recording instead. It costs 2 credits per request and needs no socket.

How#

Rules#

After the call#

How

Rules

After the call