Uploads

You can have the QA Bot score calls that did not happen in VoDialer. Upload the recordings. They get a transcript, then a review like any other call.

Do this

1. Open Quality, then Uploads. Owners, admins and quality reviewers can add recordings.
2. Click Choose recordings or drag files onto “Drop recordings here”. Pick WAV or MP3 files.
3. Optional: fill in Agent (optional), Campaign (optional) and Call date (optional). They apply to every file in this upload.
4. Click Score followed by the number of recordings.
5. Watch “This batch”. It shows how many are finished.
6. When a file shows Scored, click Open the review.

Limits

• WAV or MP3 only. The file’s content is checked, not just its name.
• Up to 50 files at once.
• Up to 100 MB for each file.

A file that is too big is refused with a message that names the limit.

Statuses

Status on screen	What is happening
Received	The file is stored
Waiting for transcription	Stored, but no transcript yet
Transcribing	The words are being written down
Transcribed	The transcript is ready
Scoring	The QA Bot is working
Scored	A review exists
Could not be scored	Something failed. The row says what

Without an AI key, files stop at “Waiting for transcription” and the page says “Transcription is waiting for AI to be switched on.” Nothing is made up. The recording stays safe and is scored once transcription works.

Two channels give better scores

The QA Bot needs to know who said what.

• A two-channel WAV (one speaker on each side) is split. The customer is channel 0 and the agent or bot is channel 1. The lines are labelled.
• Any other file, such as a mono recording or an MP3 of that kind, is transcribed whole. Its lines have no speaker labels.

When speakers are not labelled, the rule checks answer Not applicable and the judge is told. The review says “we could not tell the speakers apart”. You get a much weaker score. Ask your dialer for two-channel recordings.

Each scored upload counts as one quality review in usage. To bring recordings in by another route, see recording import.

Not yet

The transcriber has only ever run against a test stand-in. It has never run against a real AI service. The default transcription model returns text without timings, so a two-channel upload may give one long line per channel instead of a turn-by-turn conversation until the model setting is changed.

Note

The Uploads screen says files can be up to 200 MB. The server’s default limit is 100 MB, so a file between the two is refused.