Twilio Changelog

Twilio Speech Recognition with <Gather> TwiML now supports enhanced phone_call model that improves accuracy of speech recognition compared to the existing standard phone_call model. The premium phone_call model was built using thousands of hours of training data and ensures 54 percent less errors when transcribing phone conversations than the basic phone_call model. To get started, simply add enhanced=”true” attribute to <Gather> when using speechModel=”phone_call”. Click here to learn more about pricing for the premium enhanced phone_call model as well as how to use it.

An issue has been identified where Voice Insights was tagging calls as having packet loss due to Twilio-internal removal of DTMF or Comfort Noise packets.

For calls with both in-band and out-of-band DTMF key press data Twilio removes the in-band packets which results in RTP packet sequence number jumps that are interpreted as packet loss by the Voice Insights platform.

Additionally, calls where Twilio received Comfort Noise packets the CN packets are not propagated internally which results in the RTP sequence number jumps being reported packet loss.

In both of these cases the calls have been tagged as having high packet loss, but the packet loss percentage reported by Voice Insights in the call summary was 0%.

We have updated the packet loss call tagging logic to account for these conditions and these calls will no longer be tagged by Voice Insights unless the percentage loss for packets outside of these two conditions is greater than 5%.

Depending on your call flow you may see a significant decrease in the number of calls marked as having been impacted by packet loss going forward.

Following the release of Amazon Polly Neural Text-to-Speech voices for the Voice <Say> verb, Studio now supports the same new voices in the Say/Play and Gather Input widgets.

In Marketing Campaigns, you can now segment contacts from a particular list directly from the Contacts page. This reduces the steps to create list-specific segments.

You can now view WhatsApp template rejection reasons in console. When WhatsApp rejects a submitted template, they provide a reason:

• PROMOTIONAL: WhatsApp has deemed this to contain promotional or marketing content.
• TAG_CONTENT_MISMATCH: The language and/or template category selected don't match the template content.
• INVALID_FORMAT: Placeholders or other elementse formatted incorrectly.

Previously, this information was only available by opening a support ticket. For additional details on template rejection reasons from WhatsApp, please refer to our Guidance on Templates.

Happy building!

The Conference Create Participant API has been updated to ensure that any outbound call placed to a phone number uses a Twilio number or verified outgoing caller id in the From field. This change prevents malicious actors from using the API to spoof a phone call from an arbitrary phone number. Note that calls placed to a Twilio Client or a SIP address are not impacted by this change.

We are adding a new IP address block for the Media Server range in our Frankfurt, Germany (de1) region. This range will be used for the Programmable Video service and our Network Traversal Service. The new IP block is listed on the IP Addresses page in our Video documentation and on the Network Traversal Service Regions page.

If you have allowed the IP addresses of the de1 region in your Firewall, then your Firewall configuration will have to be updated to include the new IP block. This new block will go into rotation on April 16th, 2020.

Flex has supported only inbound voice flows natively so far. For outbound capability, either you had to build the Dialpad yourself, use the experimental dialpad, or use one of the community-supported plugins. With this update, Flex now offers native outbound dialing ability right from the Flex UI.

With the Flex Dialpad:

• Agents can call customers directly from Flex and associate it with a relevant queue for reporting.
• Agents can call customers while in an unavailable state, allowing them to remain uninterrupted from incoming tasks.
• While on an outbound call, agents can initiate either Warm
  Transfer or Cold Transfer to another agent or supervisor.
• Supervisors can monitor outbound calls.
• Outbound calls will be presented both in real-time and historical reporting in Flex.
• Developers can build click-to-dial use cases using the programmability offered through the new StartOutboundCall Action.

The Flex Dialpad is a part of the Flex UI and offers all of the theming and customizability features of a React component.

For more details, visit our announcement blog post and the Dialpad documentation.

You can get your hands on the Flex Dialpad now by upgrading your Flex Instance to v1.18 and then enabling Flex Dialpad in your Flex Settings in Twilio Console.

Emergency Calling is now available for Programmable Voice SIP Interfaces in the US and Canada. Customers are now able to make Emergency Calls from their IP communications infrastructure using Programmable Voice SIP Interfaces with the Twilio platform, which will route 911 calls to the appropriate Public Safety Answering Point (or PSAP) and programmatically pass along the corresponding emergency address.

Click here to learn more.

The following additions to the Video Recording service are now Generally Available:

• Recording Composition API: provides developers flexibility to mix the media files of a Room recording into a high-quality, custom layout that matches their specific use case.
• Composition Hooks API: allows developers to streamline the recording workflow by automating the creation of Recording Compositions for every Room.
• Public Key Encryption for Recordings & Compositions: for enhanced security we are offering Enterprise customers the ability to encrypt recordings and compositions with a customer provided public encryption key.
• External AWS S3 Storage: Enterprise customers can configure the Recording service to push Recordings or Compositions directly to their AWS S3 Bucket.

These Video recording features allow developers to create a recording workflow that matches their specific use case and security needs. For example, a customer in the Healthcare industry may decide to encrypt all recordings with their Public Key and have them automatically stored to their external storage, and a customer in the Education space may decide to compose the recording of a virtual classroom into a presentation layout for students.

To learn more check out the blog post or visit the documentation on Recording Composition API, Composition Hooks API, Encrypting your Stored Media and Storing to AWS S3.