How to record and transcribe a Google Meet call
2026-07-19
Google Meet only records calls on paid Workspace plans — Business Starter starts at $6 per user per month — and even then you get a video file in Google Drive rather than anything you can search or quote from. If what you actually want is the text — who agreed to what, what the client asked for, what action items were assigned — you need two things: a recording, and a transcript with speaker labels. Both are free, and neither requires a Workspace subscription.
Here is the complete workflow, from a live call to a labelled, downloadable transcript.
1. Record the call
Install Meet Voice Recorder from the Chrome Web Store. It is free and runs entirely in the browser. The extension captures the tab audio — every other participant on the call — and mixes in your own microphone, so the output is one file containing the full conversation rather than a one-sided recording.
Open your Meet call and press Ctrl+Shift+Y to start recording. The shortcut toggles: press it
again when the call ends to stop. The recording saves to your Downloads folder as
meet-YYYY-MM-DD-HHMM.webm. A typical one-hour call produces a file between 20 and 80 MB,
depending on audio quality and the number of participants.
The extension does everything locally in your browser. Nothing is uploaded to any server by the recorder itself,
and it keeps no copy. The .webm file it hands you is the only one that exists.
A few habits make for a cleaner recording. Use a headset or external microphone if you have one — the built-in laptop mic picks up keyboard noise and room echo. Mute when you are not speaking, and start the recorder before the first word is said rather than after. If you switch away from the Meet tab mid-call, the extension may capture audio from the wrong tab, so stay on it throughout.
2. Upload the file
Drop the .webm file onto the WhoSayAh home page. It is accepted as-is —
there is no need to convert it to MP3 or strip the audio first. Video containers like .webm,
.mp4 and .mkv all work because only the audio track is read. Accepted audio formats
include mp3, m4a, wav, ogg, opus, flac and aac as well.
Files up to 100 MB and 120 minutes are accepted. A rejection message appears immediately at upload time if the file exceeds either limit, so you know before any transcription time is spent. If a two-hour all-hands recording exceeds the size limit, trim it to the portion you need or split it into segments under the limit before uploading.
Once the file is uploaded, a dialog shows the file’s name, size and detected length. Two settings in that dialog matter.
Language. Leave it on Detect automatically unless the call was short or contained a lot of background noise. In those cases there is less audio to work from, and naming the language explicitly helps. For calls that switch between two languages, choose the one used for the larger portion of the conversation.
Speaker labels. Leave Label speakers switched on. Without it you get one undifferentiated block of text — for a multi-person call that is nearly useless because you can read every word and still not know who asked the question or who made the commitment. With it, each line is attributed. Enter how many people were on the call in the How many people are speaking? field. If you genuinely do not know, I don’t know — detect it starts it without a count, though entering the correct number is the single setting that most improves the result.
Press Transcribe.
3. Read as it arrives, then name the speakers
Lines appear as they are transcribed, so you can start reading before the job finishes. A 30-minute call is typically readable within 3 to 5 minutes of pressing start; longer files take proportionally longer. Each line is prefixed with Speaker 1, Speaker 2 and so on. At this stage the tool knows the voices are distinct but has no way to know whose name belongs to which.
Once the transcript is complete, press Name speakers. A panel lists each voice with two pieces of information: how long it spoke and its first line. On a call you have just been on, that is usually enough to place each person. Type in the real names and they are applied across the entire transcript instantly, replacing every Speaker N label throughout.
4. Download or copy
The transcript can be copied to the clipboard in one click, or downloaded in two formats:
- .txt — plain text, one line per speaking turn, speaker name prefixed. This is the format to paste into meeting notes, share with a colleague, or attach to a project ticket.
- .vtt — WebVTT subtitles with precise start and end timestamps on each cue. Useful if you also recorded video and want a subtitle track. The file drops straight into most video players without any conversion step.
The .vtt format is a web standard, so it is not limited to this page. It works as a subtitle
track in HTML5 <video>, on YouTube caption upload, and in video-editing software that
accepts external caption files. Teams captioning recorded meetings at scale — dozens of call recordings
a week — use the same workflow through CrocOTT,
which handles the VOD platform side and keeps captions synced with a larger catalogue automatically.
Creators who clip and repurpose recorded calls use a similar approach with
OverlayMax, which is built
for taking a recorded stream or meeting and turning it into shareable clips.
How good are the speaker labels?
Speaker attribution is reliable on a typical call where participants take turns and the audio is clear. Two things make it harder.
Very short interjections. A one-word “yeah” or “right” carries very little voice information. Those are the lines most likely to be attributed to the wrong speaker. Complete sentences are reliable; single words are not. If attribution accuracy matters for every line, recording each participant on a separate audio track eliminates the problem entirely — the two tracks can be uploaded and transcribed separately, then merged.
Heavy crosstalk. When two people speak simultaneously for more than about 2 seconds, both voices are present in the same segment and the tool has to make a judgement call. Crosstalk is unavoidable on some calls; it is worth knowing that those moments are where errors concentrate.
Enter the speaker count. This is worth repeating because it consistently has the largest impact on output quality. Left to estimate, the tool can over-split a voice — treating one person as two if they change pace or pitch noticeably between calm answers and animated ones — or merge two quiet voices into one. Getting the count right removes both failure modes at once. Even a rough estimate of 2 for a one-to-one call is better than leaving the field blank.
For more general advice on getting a clean transcript from any recording, the audio transcription guide covers microphone technique, format choice, and common recording mistakes in more detail.
A note on privacy and consent
Recording a call means recording other people. In many jurisdictions — the European Union, the UK, Canada, and around half of US states — recording without explicit consent from all participants is illegal. Even where it is not required by law, informing everyone on the call is the right thing to do and avoids disputes later.
A simple practice: at the start of the call, say “I’m going to record this for my notes — is everyone okay with that?” For formal or professional contexts, document the agreement. If a participant objects, do not record.
On WhoSayAh’s side: the uploaded file is deleted as soon as the transcript is ready. The transcript itself is deleted after 24 hours. There is no account, so nothing links either to you. The privacy policy states exactly what is stored and for how long. The terms of use list the file size and length limits in full.
For calls that cannot leave your own network — client conversations under NDA, medical or legal material, anything subject to a data-residency rule — the alternative is a self-hosted installation where no audio ever reaches an external server. FastoCloud provides the infrastructure for that use case. The workflow for the person uploading is identical to what is described above; the processing stays entirely within your own environment, and there is no per-minute charge.