Twilio Changelog | Jan. 13, 2025

Send Real-Time Transcripts to Voice Intelligence

Customers generating a Real-Time Transcript for Voice calls can now send these transcripts to Twilio Voice Intelligence for integrated post-call processing. This feature enables you to extract actionable insights at scale by storing and analyzing transcripts initially generated in real-time.

Here’s what you can do:

Persist Live Transcripts: Automatically store real-time transcriptions in Voice Intelligence’s historical log for future reference and analysis.
Run Post-Call Language Operators: Trigger configured Language Operators in Voice Intelligence to generate AI-powered insights and automate actions after the call has concluded.

This integration with Voice Intelligence can be configured either through the <Transcription> TwiML verb or through the Real-Time Transcription API. It works seamlessly alongside the existing live processing Webhook setting (statusCallbackURL), which streams utterance-level data to your app during the call.

To enable this, include the intelligenceService parameter (Voice Intelligence Service SID) when starting a transcription session. If you don’t have a Voice Intelligence Service yet, create one in Console.

Via <Transcription> TwiML verb:

<?xml version="1.0" encoding="UTF-8"?>
<Response>
    <Start>
        <Transcription 
          intelligenceService="GAaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"                    
          statusCallbackUrl="https://example.com/your-callback-url" />
    </Start>
</Response>

Via Real-Time Transcription API:

curl -X POST "https://api.twilio.com/2010-04-01/Accounts/$TWILIO_ACCOUNT_SID/Calls/CAaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa/Transcriptions.json" \
--data-urlencode "IntelligenceService=GAaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa" \
--data-urlencode "StatusCallbackUrl=https://example.com/your-callback-url" \
-u $TWILIO_ACCOUNT_SID:$TWILIO_AUTH_TOKEN

Note: To use Post-Call Language Operators with real-time transcripts, ensure these Operators are configured in your Voice Intelligence Service before starting the transcription session. Language Operators are executed after the Real-Time Transcription session concludes, either automatically through the call ending or manually by stopping the live transcription.

For more details on how to get started with Real-Time Transcription, check out our TwiML and API documentation.