Voice Agent
Get started

Upload a recording

Evaluate an existing conversation with the right channel and context.

Use Upload Recordings to evaluate real conversations that have already happened.

Before you upload

Choose the Agent under test first. The recording must include that AUT, and you should have permission to upload and evaluate it.

The current upload UI offers WAV, MP3, M4A, and FLAC files. Files are validated when processed, so a file with a listed extension still needs to be readable as audio.

The deployment defaults are up to 50 files and 500 MiB per file. Your workspace may show the visible limit as 500 MB / file; administrators can configure these defaults.

Upload a recording

  1. Open Upload Recordings.
  2. Under Select an agent, choose the AUT.
  3. Add one or more recordings.
  4. Set the AUT channel for each file:
    • Auto lets Voice Agent identify the channel.
    • L uses the left channel.
    • R uses the right channel.
  5. Select Upload & compute.

Use Auto unless you know which channel contains the AUT. Choosing the correct channel makes the transcript and metric evidence more useful.

Add context when the call needs it

Open Advanced to add optional context such as Classification, Call driver, Expected handling, Workflow checklist, Required information, and Call date (optional).

This context helps Voice Agent apply the most relevant evaluation criteria. The product automatically selects applicable metrics for uploaded recordings; you do not need to configure each one manually.

After submission

Processing continues in the background. Open Results to find the resulting suite, then open the recording to review its transcript, score, and evidence.

If a file does not process

Confirm that you selected the correct AUT and that the file is an audio format the current uploader offers. If the recording has unusual encoding or channel layout, try a known-good export before uploading again.

Understand the results

On this page