Import an audio segment for a voiceprint model

You can import audio to create a voiceprint model manually or to enhance an existing model. Uploaded and approved audio is used to model the voiceprint from a source other than the suite Recorder.

Note the following when importing audio for use by a voiceprint model:

Users with the role privilege of Approve/Unapprove Interactions for Enrollment can have uploaded audio set automatically to an approved status. A file meeting the audio requirements for enrollment contributes to the voiceprint model for the person.
After adding and approving an audio segment for an already trained voiceprint model, the Training Status changes to Invalidated. After retraining the model using approved audio segments, the status changes.
To understand the impact that imports can have on automatic enrollment, review Manual enrollment by interaction import.

Before you begin

Create a person
.
Review Audio requirements for imported files.
If you import a segment containing multiple voices, the system incorporates both voices into the voiceprint model. Multiple voices result in a corrupted voiceprint model.

Procedure

Access the Voiceprints Model for the person whose model you need to update or create.
In the Description field, enter text to identify the audio segment.
In the Audio field, click Browse. Then search for and select the audio file that you want to use to create or update (train) a voiceprint.
If uploading a stereo audio file, use the Channel field to choose to correct voice of the person speaking. Select from the following channels of the audio file:
- Stereo Left / Mono
- Stereo Right
Click Upload .

The system adds the audio to the Recorder as an interaction. The interaction start time for imported audio is the time that the import occurs.
- When a user has the role privilege of Approve/Unapprove Interactions for Enrollment, the following occurs:
  - Uploaded audio is placed automatically in an Approved state.
  - If the audio file is accepted for the voiceprint model:
    - The voiceprint model is trained using the uploaded audio segment.
    - The Training Status field changes from Untrained to Trained.
    - Training requires approximately 30–60 seconds as the audio is processed and training occurs.
- When a user does not have the role privilege of Approve/Unapprove Interactions for Enrollment, uploaded audio is placed in a Pending Approval state. Contact your manager about getting the audio approved for use by the voiceprint model.
Repeat this procedure for each audio file you want to use. Each additional file used provides for a more reliable match.

What to do next

Manually approve a voiceprint model audio segment