Your Zoom or Teams recording is too quiet — fix the saved file

Drop the file into the browser volume booster, click Smart Boost, listen, export. That's the fix, and a one-hour meeting takes about a minute of your attention. Smart Boost reads every sample in the recording, works out how much headroom is sitting unused, and tells you the exact percentage that makes the file as loud as it can go without a single clipped sample. On a quiet Zoom or Teams recording that number usually lands between 200 % and 500 %.

One thing to get out of the way first: no browser tool can turn up your computer's system volume, and no iOS app can either. What this does is rewrite the audio itself, so the file is louder everywhere you play it afterwards.

The fix, step by step

1. Open the volume booster and drag your recording in. A Zoom local recording gives you both a .mp4 and an audio_only.m4a; use the m4a if you only need audio. Teams recordings land in OneDrive or SharePoint as .mp4, so download that first.
2. Click Smart Boost. You get four numbers back: sample peak in dBFS, RMS in dBFS, integrated loudness in LUFS, and headroom in dB. A file peaking at -12 dBFS has 12 dB of headroom. Smart Boost keeps 1 dB of margin, so it will suggest roughly +11 dB, which is 355 %.
3. Apply the suggested percentage and leave the soft limiter on. It's on by default and it does nothing at all below 0.7 of full scale.
4. Hit A/B compare a few times. It level-matches the two sides, so you're judging whether the voices got clearer instead of just louder.
5. Export as WAV. 16-bit is plenty for speech: about 10.6 MB per minute at 44.1 kHz stereo. Use 24-bit if an editor is going to work on it afterwards.

The slider stops at 500 %, which is +14 dB. If your file peaks quieter than about -15 dBFS, one pass won't get all the way there. Export the WAV, drop it back into the same page, and run Smart Boost again on the result.

Why Zoom and Teams recordings come out so quiet

It's rarely one cause. Usually it's three stacked on top of each other.

Automatic gain control on every mic. Both platforms run AGC on each incoming participant so nobody blows out the call. AGC is tuned for live intelligibility, not for something you'll listen back to later, and it's conservative. It pulls loud talkers down more eagerly than it pushes quiet ones up.

The recording captures the mix, not the mics. A local or cloud recording is the summed conference stream, the same thing everyone on the call heard. That mix is deliberately kept well below full scale so two people talking at once doesn't distort. Even a clean meeting recording often peaks around -12 to -20 dBFS with integrated loudness near -30 LUFS. Streaming platforms target -14 LUFS. You're 16 LU down before anything actually went wrong.

One quiet participant drags the file down. Somebody on a laptop mic two metres away, or a speakerphone in a conference room with people spread around the table, arrives 15 to 25 dB below the person wearing a headset. You end up riding your own volume up for them and back down for everyone else. That's what "our meeting recording is barely audible" almost always means: the file isn't uniformly quiet, it's unevenly quiet.

Room audio makes it worse in a way that gain can't fix. A speakerphone picks up the room's reflections along with the voices, so raising the level raises the reverb with it. You can improve that recording. You can't undo the room.

Don't upload a meeting recording to a random site

Most "online audio booster" pages work by sending your file to a server, processing it there, and handing back a download link. For a music clip, fine. For a meeting recording, think about what's in the file: client names, pricing, unreleased plans, a performance conversation, an HR complaint, someone explaining a medical absence. It also contains other people's voices, which in two-party-consent jurisdictions isn't yours to hand around.

These tools run entirely inside the browser tab. Your browser decodes the file, your machine processes it, your machine writes the result back out. No server, no account, no upload. If you want proof rather than a promise: load the page, switch off Wi-Fi, then drop the file in. Everything still works, because at that point nothing needs the network.

Loud is not the same as clear

Once the level is right, some recordings are still hard to follow. That's tone, not gain, and the five-band equalizer is where you fix it. The Voice preset sets the bands like this:

The two cuts at the bottom pull out air-conditioning rumble, laptop fan noise and desk thumps, all of which eat headroom without carrying any speech. The +5 dB at 3 kHz does the real work. That's the presence region where consonant transients live, and it's the difference between hearing a word and guessing it from context.

Be careful with the top band. Meeting recordings are heavily compressed audio, and there's often very little real content left up at 14 kHz. Boosting it can just raise codec hiss. Try it, A/B it, and back it off if all you gained was noise.

These are peaking biquads at Q = 1.0, so each one is about 1.4 octaves wide, roughly two-thirds of an octave either side of its centre frequency. The live response curve shows you exactly what you've built.

If the recording sounds boomy rather than dull, which is common with a conference-room speakerphone, do the opposite and cut. Leave the bass booster alone entirely for this job. Its 120 Hz low shelf is built for music, and on a meeting file it mostly lifts rumble.

Use the level-matched A/B after every change. It trims the processed side back to the original's peak, so you're comparing tone against tone. Without that, louder always sounds better, and you will happily EQ yourself into a worse file.

When one speaker is loud and another is quiet

This is the case where more gain actively hurts. If your loudest talker peaks at -6 dBFS and your quietest sits around -28 dBFS, Smart Boost can only offer about +5 dB before the loud one hits the ceiling, and the quiet one is still inaudible.

So push past the Smart Boost number and watch the live clip warning, which tells you what percentage of samples would run past full scale at your current setting. At 0.1 % on a speech recording you'll hear nothing wrong. The limiter's tanh knee is transparent below 0.7 of full scale and bends asymptotically toward the ceiling above it, so loud syllables get rounded off rather than chopped square. Somewhere around 3 to 5 % you'll start hearing the loud talker sound pinched. Back off from there.

The limiter is website-only. The iPhone app and the Android app share the same five EQ bands and the same preset values and go to 200 % volume, but neither has the limiter or the level-matched A/B. For an uneven meeting file, do it in the browser.

An honest limit: this is a limiter, not a per-speaker leveller. There's no attack or release, and it doesn't ride the gain across minutes the way a broadcast processor would. If the imbalance is severe, the real fix is upstream. Zoom can record a separate audio file per participant, but the setting has to be switched on before the meeting starts. Nothing you do afterwards beats that.

What you get back: WAV, and the video is untouched

Export is WAV only, at 16-bit, 24-bit or 32-bit float. There is no MP3 encoder here. A 90-minute meeting as 16-bit 44.1 kHz stereo runs about 950 MB, which is fine for handing to an editor and hopeless for email. Convert it afterwards if you need something small, or trim to the section people actually need before you export.

Rendering runs offline rather than in real time, so it finishes in a fraction of the recording's running length. The one hard limit is memory, because the whole decoded file lives in RAM while you work on it. Multi-hour recordings on an 8 GB machine can get tight. Split them if the tab starts struggling.

How loud should the finished file be?

Smart Boost measures integrated loudness with a real ITU-R BS.1770-4 implementation: K-weighting, 400 ms blocks at 75 % overlap, an absolute gate at -70 LUFS and a relative gate at -10 LU. That's the same measurement broadcasters and streaming platforms use, so the number compares directly to their targets.

The tool shows you the gap and suggests a gain. It doesn't quietly re-master the file to a target behind your back. You read the number and move the slider, or you click the suggestion.

Two things worth knowing about that measurement. It reports sample peak, not true peak (dBTP), so a file sitting at exactly 0.0 dBFS sample peak can still overshoot a little once it's encoded to a lossy format or reconstructed by a converter. That's why the zero-clipping suggestion leaves 1 dB of margin. And DRM-protected audio can't be decoded in a browser at all, so a recording locked inside a protected player is out of reach for any tool of this kind, this one included.