Voice | Sep. 22, 2026
Introducing Provider Failover and “Auto” mode Provider selection for Real-Time Transcriptions on October 22, 2026
Starting October 22, 2026, Twilio Real-Time Transcriptions automatically starts transcription sessions on a second speech provider when a customer’s primary provider doesn’t connect. Failover is enabled for all existing Real-Time Transcriptions accounts on that date, and it routes only within a customer’s existing region. Use of failover is optional and can be opted out of per sub-account in Console.
We’re also launching “Auto” mode on October 22, which will become the default TranscriptionEngine selection for new Real-Time Transcriptions customers, which means Twilio selects the best speech provider to use per session based on criteria such as Word Error Rate for that language, availability, and more, instead of sending every session to a named provider.
Existing customers’ configurations of Transcription Provider / transcriptionEngine are not changing. To opt in to Auto mode as an existing Real-Time Transcriptions customer, set transcriptionEngine=auto on the <Transcription> TwiML noun or on Calls Transcriptions API requests.
What’s changing with Real-Time Transcriptions?
Until now, a transcription session depended on a single speech provider being reachable, and a session that could not connect returned an error. Twilio now retries the alternate provider instead, so a transcription starts on Deepgram if Google does not connect, and on Google if Deepgram does not connect. Customers get multi-provider resiliency with no custom routing or payload normalization of their own.
Here are more details on how provider failover works:
Failover triggers at session startup only, not mid-session, and it stays inside the customers’ existing region, so a session processed in Europe fails over to the alternate provider in Europe.
If the alternate provider does not support profanity filtering, custom hints, an exact language variant, or provider word timing data, Twilio drops or downgrades that setting and returns a working transcript instead of an error.
The setting applies as normal on the next session with the usual provider. Sessions that use language=multi are unchanged and continue to run only on Deepgram’s Nova-3 multilingual model.
Need a hand?
These resources will help: