How to Transcribe a Phone or Conference Call Recording
Turning a recorded phone call into text: getting the recording as a file, why phone audio transcribes a little worse than a room recording, and what to do about it.
The Scriptorium team ·
Get the call recording as an audio file, then upload it to audio to text the same as any other recording. The one thing worth knowing going in: phone audio is narrowband, meaning it carries a smaller range of frequencies than a room recording, so a phone call transcribes reasonably well but usually a little worse than the same conversation recorded in person.
Getting the recording out as a file
This depends entirely on how the call was recorded, since transcription tools work from a file and cannot reach into a call in progress:
- A call-recording app (phone-side) typically saves each call as an audio file in its own app, exportable via share or export.
- Conferencing software (dial-in conference lines, not video meetings — see the dedicated Zoom, Teams and Google Meet articles for those) usually offers a downloadable recording from the host's account after the call ends.
- A carrier or PBX voicemail/recording feature generally provides a download link or attaches the recording to an email.
Whatever the source, once it is a standalone MP3, WAV, M4A or similar file on your device, it behaves exactly like any other audio recording from here.
Why phone audio transcribes a little worse
Landline and most cellular calls carry a narrower band of frequencies than a normal room recording — a deliberate trade-off phone networks make to save bandwidth, going back to analogue telephony. That discards some of the acoustic detail a speech recognizer would otherwise use, particularly for sibilant sounds ("s", "f") that sound similar to each other on narrowband audio. It still works; it is just not the same starting point as a clear in-person recording.
Calls made over a data connection through an app (rather than a traditional phone line) sometimes carry more frequency range than a standard cellular call and can transcribe a little better as a result — but this varies by app and connection quality, and is not something to rely on.
Multiple people on the call
Conference calls with several participants benefit the most from speaker labelling, since a phone-call transcript with no attribution is harder to follow than a video-meeting transcript where you at least remember who was on screen. Speech is grouped by speaker and each turn labelled, up to eight speakers, though a phone line's audio quality makes distinguishing very similar-sounding voices somewhat harder than it is on a clean room recording.
If the call was two-way and you only have one side
Some recording setups (certain call-recording apps, one-sided consent recording) capture only your own audio, not the other party's. There is no way to recover audio that was never captured — the transcript will only contain what is actually in the file, half a conversation and all.