Product changelog and announcements

TL;DR: Twilio’s Real-Time Transcriptions product has added Deepgram Nova-3 Multi-language speech model support, now in Public Beta,  for use with either Webhook-delievered speech results, or with Persisted Transcript Resources and Twilio’s Conversational Intelligence.

 

What are Twilio Real-Time Transcriptions?
Twilio Real-Time Transcriptions allows you to transcribe live calls in real-time. When Twilio executes the <Start><Transcription> instruction during a call, the Twilio platform forks the raw audio stream to a speech-to-text Transcription Engine, which then provides streamed speech recognition responses back with each of the caller’s uttered phrases. Developers can choose to send that stream of speech recognition results to their downstream app through Twilio Programmable Voice, either using webhooks, or they can opt instead to  send them to a configured persisted transcript resource on the Twilio Platform – and, in either case, Developers can opt to use either Google or Deepgram as the Transcription Engine providing those transcribed speech results.

 

What are the New Multi-Language-Detecting Capabilities of Real-Time Transcriptions?
Now as a Public Beta capability of Real-time Transcriptions, when Developers opt to use the Deepgram Transcription Engine and opt for Deepgram’s Nova-3 speech model, they can also select “Language = multi” –  and then Deepgram’s multi-lingual Nova-3 speech model will detect (and programmatically return) the languages being spoken on the call from among the 10-plus languages that Deepgram’s Nova-3 Multi-language model supports today, and will also transcribe all the speech on the call in those (multiple) languages detected, as text in each of those detected (multiple) languages. 

 

Customer benefits 
With the streaming speech recognition capabilities of <Start><Transcription>, businesses can capture the full text of what all their customers are saying – whether to a human agent or an automated self-service AI agent or LLM – now, even across a mixed set of (multiple) languages, with this Public Beta language detection feature. 

Multi-Language Detection and transcribing  is perfect for capturing conversations: 

•  where multi-lingual agents may be speaking any one of a set of languages they are fluent in with customers, but with an individual customers’ language not known apriori,

• where the caller themselves mix languages – such as using a mix of Spanish and English, switching back and forth multiple times during the call depending upon caller language comfort and word complexity –  and then in adding that all that speech data accurately to a caller’s customer record, be that in a CRM or another application/system built by the developer, and

• where customer data collection via programmable outbound calling is the objective, either for follow-up, post-service, or post-care surveys, etc – no matter what combination of languages may be used in the prompting or in the called parties’ responses.
  

Twilio Real-Time Transcriptions allow developers to automate the capturing of customer speech data, programmatically, for each and every call (instead just having the data for an ad hoc sampling of calls), create a repository of structured data for those voice conversations with customers, and easily and cost-effectively stream the speech results to downstream applications during calls with customers.

 

More information: 

https://www.twilio.com/en-us/speech-recognition

https://www.twilio.com/docs/voice/twiml/transcription#speechmodel

https://www.twilio.com/docs/voice/api/realtime-transcription-resource

https://www.twilio.com/docs/conversational-intelligence

https://www.twilio.com/en-us/voice/pricing/us (See “Conversational Intelligence - Transcription, Streaming (Real-Time) Transcription)

TL;DR: Twilio’s Real-Time Transcriptions product has added Deepgram Nova-3 (monolingual) support for use with Persisted Transcript Resources, now HIPAA eligible!  Now your real-time transcripts capturing what customers say can also be analyzed by Twilio’s Conversational Intelligence’s Language Operators! Start transcribing using just a simple <Start><Transcription> TwiML instruction or API call, and setting up the IntelligenceService attribute there.

 

What are Twilio Real-Time Transcriptions?
Twilio Real-Time Transcriptions allows you to transcribe live calls in real-time. When Twilio executes the <Start><Transcription> instruction during a call, the Twilio platform forks the raw audio stream to the speech-to-text transcription engine, which can provide streamed responses back with each of the caller’s uttered phrases. Developers can choose to send the stream of speech recognition results to their downstream app through Twilio Programmable Voice, using either webhooks (as GA’d previously) – or send them to a configured persisted transcript resource on the Twilio Platform. With persisted transcript resources for Real-Time Transcripts, developers can opt either to use Google as the Transcription Engine, or now, use Deepgram (GA’d as of today) – both which can also now be used in conjunction there with Twilio’s Conversational Intelligence capabilities to analyze the transcript, post-call. 

 

What are the new features of Real-Time Transcriptions?
We’ve added options! Now, in Real-time Transcriptions, Deepgram’s next generation speech model Nova-3, in monolingual variants – in addition to the Deepgram Nova-2 speech models already supported -- can be used, with or without hints, and using persisted transcript resources as the means of receiving transcript results, instead of just webhooks, if a developer so desires. 

Additionally, Real-Time Transcriptions using either persisted transcript resources or webhooks are now a HIPAA Eligible Service, to safeguard customer interactions regarding health information in sessions that the Twilio platform transcribes. 

 

Customer benefits 
With the streaming speech recognition capabilities of <Start><Transcription>, businesses can capture the full text of what all their customers are saying – whether to a human agent or an automated self-service AI agent or LLM – for doing any of the following (and more): 

• Capturing crucial customer conversations, and adding that data to a caller’s customer record, be that in a CRM or another application/system built by the developer.

• Analyzing caller-agent interactions, for near real-time escalation to supervisors, prompting for upsells, or other taking other interventional or incremental steps with the customer, while they are still on the phone.

• After sending the caller’s transcribed speech to an AI Agent / LLM, coming back to prompt a human agent with recommended actions or requested product information based on what the caller has said.

• Automating customer data collection via programmable outbound calling applications, for follow-up, post-service, or post-care surveys, etc.

Twilio Real-Time Transcriptions allows developers to automate the capturing of customer speech data, programmatically,  for each and every call (instead just having the data for an ad hoc sampling of calls), create a repository of structured data for those voice conversations with customers, and easily and cost-effectively stream the speech results to downstream applications during calls with customers.

 

More Information: 

https://www.twilio.com/en-us/speech-recognition

https://www.twilio.com/docs/voice/twiml/transcription

https://www.twilio.com/docs/voice/api/realtime-transcription-resource

https://www.twilio.com/docs/conversational-intelligence

https://www.twilio.com/en-us/voice/pricing/us (See “Conversational Intelligence - Transcription, Streaming (Real-Time) Transcription)

We’ve added the Preflight API to the Twilio Programmable Voice React Native SDK, bringing the same pre-call diagnostics available in the native iOS and Android SDKs to React Native apps. Preflight provides a pre-call assessment of network stability, bandwidth, latency, jitter, MOS scores, and packet loss. Use it to surface network warnings, prioritize mobile data, or block calls on unreliable networks, improving connection success and call quality.

Check out the docs for integrating Preflight into your Voice application. We can’t wait to see what you build!

Segment is changing how we process incoming data for profiles that exceed their configured identifier cardinality limits.

Previously, profiles that surpassed an identifier limit (for example, more than 5 emails per profile) were marked as frozen and all subsequent data for those profiles—including events and traits—was dropped.

Starting in October 2025, Segment will continue to drop identifiers for types that have already exceeded their configured limit, but will resume accepting event and trait data, as well as new identifiers that remain within their limit. This change helps ensure that profiles that were previously paused due to limit violations continue to stay up-to-date with new behavioral and trait data.

Impact:

• Audience sizes may increase or decrease depending on event or computed-trait membership criteria.

• Some profiles may enter or exit active journey steps following this update.
  

No action is required on your part. If you have questions about how this change may affect your audiences or data flows, please contact friends@segment.com for assistance.

WhatsApp typing indicators are now supported on Twilio. Create typing indicators to signal to WhatsApp users that a response is being prepared. This improves user experience by reducing perceived wait times. Use typing indicators for responses that might take longer than a few seconds to generate.

To learn more about typing indicators on WhatsApp and how to send them, see our docs here.

Twilio SendGrid now requires HTTPS for all new Event Webhook configurations. Existing HTTP webhooks must be updated to HTTPS by February 11, 2026 to continue receiving event data.

We have added support for automatically detecting the language spoken by the user or sent via a text token for ConversationRelay.

We have added support for SSML tags to ConversationRelay to provide the pronunciation of a word or an acronym, specify where pauses should be, or increase or decrease the speed of spoken text.

Compliance Toolkit now suppports U.S. state-specific quiet hours, automatically rescheduling non-essential messages sent during select time windows to help you stay compliant with specific state regulations.

This release covers select quiet hour restrictions in the following states: Alabama, Connecticut, Florida, Louisiana, Maryland, Mississippi, Oklahoma, Tennessee, Texas, and Washington.

Learn more and see how to enable it in our docs here.