Guides · August 26, 2026
How to record a meeting for a transcript
Transcript quality is decided before you press record. Microphone placement, room noise, formats, and the settings that matter, so free transcription with speaker labels actually works on your audio.
Almost everything about transcript quality is decided before you press record. People tend to blame the transcription tool for a bad result, but no speech model can recover detail the microphone never captured. Ten seconds of thought about where the recorder sits will do more for your transcript than any setting in any app.
This is a practical guide to getting audio that transcribes cleanly, whether you are recording in a room, on a call, or on your phone.
The one thing that matters most
Microphone distance. Sound falls off fast with distance, so a recorder twice as far away picks up roughly a quarter of the energy from the speaker while picking up the same amount of room noise. That ratio, not the price of the microphone, is what determines whether the words are recoverable.
In practice this means putting the recorder in the middle of the table rather than in front of yourself, and moving it closer to whoever is quietest rather than whoever is loudest. If one person is across a large table, the transcript will show it.
Even distance matters for a second reason: automatic speaker separation groups voices partly by how they arrive at the microphone. When one person is loud and close and another is quiet and far, the labels get confused far more often.
The room
Hard surfaces reflect sound, and those reflections arrive a few milliseconds after the direct sound. To your ear that is barely noticeable. To a speech model it smears the boundaries between words. A carpeted room with soft furniture transcribes noticeably better than a glass-walled meeting room with a long table.
Things worth doing, in order of payoff:
- Turn off steady noise sources. Air conditioning, fans, and projectors sit right in the frequency range of speech. A fan you stop noticing after a minute is still there in the recording.
- Close the door. Corridor noise is intermittent, which is worse than steady noise because it is unpredictable.
- Avoid cafes. Espresso machines, music, and background chatter are the hardest combination there is. Background speech is particularly bad, because the model cannot tell it apart from your speech.
- Move the recorder off the table surface if people tap, write, or set cups down. A folded jacket under it absorbs a surprising amount of thump.
Formats and settings
Recording settings matter less than placement, but the defaults on some devices are worse than they need to be.
- Record WAV if you have the option and the storage. It is uncompressed, so nothing is thrown away. An hour is roughly 300 MB at CD quality, which is nothing on a modern phone or laptop.
- If you record compressed, do not go below about 128 kbps for speech. Very low bitrate recordings lose exactly the high-frequency detail that distinguishes similar consonants. Re-encoding a low bitrate file to a higher one recovers nothing, so this choice is permanent.
- Mono is fine. Speech transcription works from a single channel anyway, so stereo mostly doubles your file size. The exception is a setup with one microphone per person, where separate channels are genuinely useful.
- Sample rate above 16 kHz gains you nothing for transcription. Most models work at 16 kHz, so a 48 kHz recording is downsampled before processing. Record higher if you want the audio for other purposes, but do not expect a better transcript from it.
- Avoid 3GP. Some older Android call recorders default to it, and no browser can decode it, so you will have to convert before you can transcribe. AMR, the other format those recorders use, works in MeetingScribe directly because the tool carries its own decoder, but its 8 kHz phone-line quality still makes transcription harder than M4A or MP3.
Remote meetings
Video calls have a specific advantage worth exploiting: the platform can often record each participant separately.
In Zoom, a local recording option saves a separate audio file per participant. That eliminates the speaker separation problem entirely, because you already know whose file is whose. You transcribe each one and interleave the results by timestamp. It is more work, and it is worth it for interviews, disciplinary meetings, and anything where misattributing a sentence would be a real problem. Details are in our guide to transcribing Zoom recordings.
Otherwise, remote audio has one quality trap: it is compressed by the platform in real time and then compressed again when saved. That double compression is why call recordings often transcribe worse than in-person ones despite sounding fine live. Ask people to use headsets rather than laptop speakers and microphones, which also removes the echo that speakerphone setups create.
Recording on a phone
A modern phone is a genuinely good recorder, and it is usually the one you actually have. Two habits make the difference:
- Know where the microphone is and do not cover it. On most phones it is at the bottom edge. Laying the phone face down on a table often points it at the surface.
- Put it in airplane mode. Not for privacy, but because notification vibrations against a hard table are loud enough to obliterate a few words each time.
Test before you rely on it
The most common recording disaster is discovering afterward that the file is empty, the microphone was muted, or the app stopped when the screen locked. Record fifteen seconds, play it back, and confirm you can hear it. Do this every time with an unfamiliar setup.
It is also worth saying the date, the purpose, and who is present in the first few seconds. That information survives file renaming, and it lands at the top of your transcript where it is most useful.
Length
If you are transcribing in a browser, plan around a practical ceiling of roughly an hour per file. Transcription itself is chunked automatically, but the whole recording is decoded into memory before that starts, and that is what sets the limit. For longer meetings, either record in segments or split the file afterward into parts named in order, then select them together so they join into one transcript.
Consent
None of this matters if you should not be recording in the first place. Recording and consent laws vary by location, and some places require every participant to agree. Our plain-English guide to recording law is general background, not advice for a specific situation. Beyond the legal question, telling people they are being recorded is usually the right call anyway, and it tends to produce a better meeting than a recording nobody knew about.
The short checklist
- Recorder in the middle, closer to the quietest person.
- Fans and air conditioning off, door closed.
- WAV if possible, never below 128 kbps if compressed.
- Headsets on remote calls, per-participant recording if available.
- Phone in airplane mode, microphone unobstructed.
- Test fifteen seconds and play it back.
- State the date and who is present at the start.
- Tell people they are being recorded.
Get these right and the transcription step becomes the easy part. When you are ready, the transcriber runs in your browser and your audio never leaves your device.
Try it yourself, free and private
Transcribe audio in your browser. Your audio never leaves your device.
Open the transcriber