Public Beta Caveats and Limitations
Public Beta
Voice Intelligence is currently available as a public beta release. Some features are not yet implemented and others may be changed before the product is declared as Generally Available. Beta products are not covered by a Twilio SLA.
Learn more about beta product support.
General Limitations and Caveats
- As a Public Beta product, Voice Intelligence is currently not included within Twilio’s helper libraries. You will need to make requests directly against the API while the product is in Public Beta.
- Voice Intelligence is not PCI or HIPAA compliant.
- Encrypted recordings are not supported.
Transcripts App
- In past pre-release documentation, the Transcripts App was called Discovery. The old name is still used in some places, such as app retrieval URLs.
- Transcripts are indexed and available for search for 90 days.
- Search terms are linked with the OR operator. For example,
macaroni salad
will match transcripts that havemacaroni
ORsalad
).
Transcription
- The following limitations apply when fetching media stored in a third-party location (specified by a
MediaUrl
).- Currently, Voice Intelligence does not send an
X-Twilio-Signature
header for media fetch requests. As a result, media stored in Twilio Assets needs to be made public. - If you use external recordings, Basic authentication on
MediaUrl
s is not supported. If you store the recordings on S3, use a presigned URL. And when storing them on Azure Blob Storage, use a Shared Access Signature (SAS). MediaUrl
s that respond with a HTTP status code that is not a 200 will result in a transcription request that does not complete.- Requests to fetch third-party media are performed exactly once. There is currently no retry behavior.
- Currently, Voice Intelligence does not send an
- Voice Intelligence does not perform speaker diarization on recordings. The use of mono recordings will result in lower transcription accuracy.
- Recordings that are under two seconds in length are not transcribed.
- Only up to two Participants can be overridden in the Channel object of the Transcript resource.
- Voice Intelligence will only process recordings or 3rd party media that are under 1GB.
Language Operators
- Pre-built Language Operators cannot be run on a specific channel of a recording.
Transcription Viewer
- In past pre-release documentation, Transcription Viewer was called Annotator or Conversation Viewer. The old name is still used in some places, such as app retrieval URLs.