Optimal Speaker Separation helps with the creation of transcripts and call insights, so we recommend that orgs enable the feature to get the most out of Einstein Conversation Insights.
Required Editions
| Available in: Lightning Experience |
| Available in Einstein Conversation Insights, which is available in Starter, Pro, Enterprise, Performance, Unlimited, and Agentforce 1 Sales Editions |
Most video and voice call providers identify speakers on the call. However, in cases where this information isn’t identified or shared, the Optimal Speaker Separation feature is used to help in the creation of transcripts and call insights. This includes non-English calls, video calls where the transcript is not shared, and voice mono calls (those that only use one channel for transmission.
Many Einstein Conversation Insights features rely on the availability of separate speaker tracks, so we recommend that all users enable the feature. If a user opts out from Optimal Speaker Separation and makes a call that would otherwise use the feature, then some call insights are still available, but no transcript or actions items are generated.
Acoustic data used for speaker separation is private and is not tracked or used for purposes other than speaker separation in features that require it. Speakers are separated but they are not identified. After we process calls for transcription, conversation insights, action items, and coaching recommendations relevant to your teams:
- We delete the acoustic features that are used to separate speakers from our system.
- The data is not retained after an individual call is processed.
- Speaker data is not shared with third parties.
- The insights generated are private to your org.