Skip to content
Call recordings

Call recording transcription for automatic QA

Merivex turns call recordings into evaluated conversations. Upload up to 50 recordings at once: each call is transcribed, mapped by the Call Transcription Agent and scored against your own scorecard, exactly like a chat or an email. It is built for contact centers and BPOs that record every call but only have time to review a few.

Step by step

How do you QA call recordings automatically?

Four steps, from audio file to a scored evaluation with cited evidence. Nothing to install.

  1. Upload the recordings

    In Data sources, open Call recordings and drag in up to 50 files at once. Three are processed in parallel, each with its own progress and status. MP3, WAV, M4A, OGG, FLAC and WebM are accepted, up to 50 MB and 2 hours per call. The call time is read from the file name.

  2. Transcribe each call

    Each recording is transcribed by Whisper-based speech recognition, with an automatic fallback if the first attempt fails. English and French are supported. The transcript becomes an interaction, counted once like any other conversation.

  3. Map the call

    The Call Transcription Agent reads the transcript and fills in the customer, the human agent, the topic, a one-sentence summary and the contact reason. It never overwrites what you entered and never creates people: the speaking agent is matched to exactly one existing team member, or left empty.

  4. Evaluate it

    The call is then evaluated automatically against your scorecard: a score per criterion, a rationale and a citation to the transcript turn, checked by the verifier agent. Compliance breaches go to human review on every plan.

The specifics

What are the formats and limits?

Six audio formats

MP3, WAV, M4A, OGG, FLAC and WebM.

50 files at once

Drag and drop up to 50 recordings in one go. Three are processed in parallel.

50 MB and 2 hours per call

A file over either limit is refused with the reason shown, never cut short silently.

Audio kept 90 days

Recordings are deleted after 90 days. Transcripts and evaluations remain.

English and French

Speech recognition runs with English and French language hints.

Encrypted, direct upload

Files go straight to encrypted object storage over short-lived signed links. Nothing passes through anyone's inbox.

Call Transcription Agent

What does the Call Transcription Agent do?

The 15th Merivex AI agent. It runs after each transcript is ready and before the evaluation.

Fills in the call

Customer, human agent, topic, a one-sentence summary and the contact reason (intent), taken only from what the transcript states.

Never overwrites

Anything the uploader entered is kept as it is. The agent only fills empty fields.

Never creates people

The speaking agent is matched to exactly one existing team member. No match, no assignment.

Never blocks the evaluation

If mapping fails, the call is still evaluated. Switch the agent off and mapping stops; transcription and evaluation continue.

Questions

Call recordings, answered

Can Merivex evaluate call recordings?

Yes. Upload MP3, WAV, M4A, OGG, FLAC or WebM recordings, up to 50 files at once and up to 50 MB and 2 hours per call. Merivex transcribes each call, maps it with the Call Transcription Agent and evaluates it against your scorecard like any chat or email.

How long is call audio kept?

Audio is kept for 90 days, then deleted. The transcript and the evaluations derived from it remain, under your plan's history window: 12 months on Starter, 24 months on Growth, unlimited on Enterprise.

Does the Call Transcription Agent change what I entered?

No. It only fills fields that are empty: customer, human agent, topic, summary and contact reason. It never overwrites what the uploader entered and never creates a person; the agent on the call is matched to exactly one existing team member or left unassigned.

Can I import transcripts from my telephony platform instead?

Yes. Genesys Cloud transcripts come from Speech & Text Analytics, Twilio call transcripts from Voice Intelligence, Aircall transcripts from Aircall AI, and Talkdesk provides its own. Five9 reports carry no transcripts, so Five9 calls arrive as call records without text.

Does Merivex analyze tone of voice?

No. Calls are scored on what was said, from the transcript. There is no acoustic analysis of tone, silence, overlap or talk ratio, and Merivex is not a real-time agent-assist tool: it evaluates calls after they happen.

Which plans include call recording transcription?

Every plan. Each transcribed call counts as one interaction toward your monthly allowance, like any other conversation.

Keep reading

Score last week's calls

Drop a batch of recordings, let them transcribe, and compare what Merivex finds with what your sample told you.