Voice memo too quiet — how to rescue the recording you already made

A faint voice memo is almost always fixable. The recording isn't damaged, it's just quiet: the loudest moment sits far below the point where digital audio runs out of room, and all that empty space is free volume. On a phone recording of someone talking a few feet away, the gap is usually 15 to 25 dB. Adding 15 dB makes the waveform about 5.6 times taller; 25 dB makes it roughly 18 times taller. Neither costs you any quality, because you're filling space that was empty. Drop the file into the browser volume booster, press Smart Boost, and it tells you the exact percentage that uses that space without clipping a single sample.

What "headroom" actually means on your file

Digital audio has a hard ceiling called full scale, written 0 dBFS. Nothing can be louder. A sample that tries gets flattened at the top, which is the crunchy distortion people are scared of. Everything quieter is a negative number. A loud pop song peaks around -1 dBFS. A voice memo recorded across a meeting table often peaks at -20 dBFS, and one recorded in a pocket or at the back of a lecture hall can peak at -35 dBFS.

The distance between your file's peak and 0 dBFS is the headroom. When Smart Boost scans the memo it reports:

• Sample peak, in dBFS — the single loudest sample in the file.
• RMS, in dBFS — average energy, which tracks how loud something feels better than the peak does.
• Integrated loudness, in LUFS — a real ITU-R BS.1770-4 measurement: K-weighting, 400 ms blocks at 75 % overlap, absolute gate at -70 LUFS and relative gate at -10 LU. This is the number streaming platforms use.
• Headroom, in dB — plus the exact boost percentage that leaves 1 dB of margin and produces zero clipped samples.

One click applies that suggestion, and because the maths is done per sample rather than guessed, nothing distorts. A file peaking at -13 dBFS gets a suggestion near 400 %, since 400 % is +12 dB.

Where the 500 % ceiling bites

The booster runs from 0 to 500 %, and 500 % is +14 dB. That covers most memos. It does not cover a recording that peaks at -35 dBFS, which needs about 34 dB to get near full scale.

There's a clean way around it. Boost by the full amount, export as a 32-bit float WAV, load that file back into the tool and boost again. Two passes get you 28 dB, three get you 42 dB. Float WAV has no fixed noise floor and no quantisation step, so the extra round trip costs you nothing measurable. Don't do this with 16-bit exports, where each pass does add a little quantisation noise.

What no number of passes fixes is the noise underneath.

The honest limit: the noise comes up too

Gain is dumb. It multiplies everything in the file by the same amount. The voice gets 18 times louder and so does the air conditioning, the traffic outside, the hiss of the phone's own preamp, and the person shifting in their chair. The signal-to-noise ratio doesn't change at all. You moved the whole thing up.

In practice: a memo where the speaker was close to the phone comes back sounding genuinely good. A memo recorded from across a room comes back audible but hissy, and past a point more gain stops helping because you're mostly amplifying the room. These tools do not do noise reduction, spectral repair, or speech isolation. If a recording is buried under crowd noise, boosting won't dig it out, and any browser tab promising that is overselling.

Most "voice memo too quiet to hear" cases aren't a noise problem, though. They're a level problem, and a level problem has a clean fix.

Use the Voice EQ before you reach for more gain

Speech intelligibility lives mostly between 2 kHz and 5 kHz. That's where consonants are, and consonants are the difference between "fifteen" and "fifty". Low frequencies carry rumble, handling noise, HVAC and the boomy proximity effect, and none of that helps you understand a word. Raising the whole file lifts the useless part exactly as much as the useful part.

Cut the bottom, lift the presence, and the memo reads clearly at a lower overall level than brute gain would need. The five-band equalizer has a Voice preset that does precisely this:

These are peaking filters at Q = 1.0, so each one works over a bit more than an octave around its centre frequency. The -4 dB at 60 Hz throws away rumble you were never going to hear on a phone speaker. The +5 dB at 3 kHz does most of the work.

Two adjustments worth making by hand after you load the preset. If the speaker sounds boxy, pull 230 Hz down further, to -4 or -5 dB. If the sibilance turns harsh, back 3 kHz off to +3 dB and take 14 kHz to 0 — on a noisy memo that top band mostly amplifies hiss.

Use the level-matched A/B button while you do it. It trims the processed signal back to the original's peak before switching, so you hear the tone change instead of just "louder". Louder always sounds better for the first two seconds, which is exactly why untrimmed comparisons mislead you.

Step by step, start to finish

1. Get the file onto the device you're working on. On iPhone: open Voice Memos, tap the recording, tap the three dots, choose Share, then Save to Files. AirDrop to a Mac works too, and so does emailing it to yourself. Voice Memos exports .m4a, which Safari, Chrome and Edge decode without complaint; Firefox depends on the codecs your OS has installed.
2. Open the volume booster and load the file. Nothing is uploaded and there is no server — the audio is decoded in your browser and stays on your device. Turn Wi-Fi off after the page loads and the tool still works.
3. Press Smart Boost. Read the headroom figure, then apply the suggested percentage. Leave the soft limiter on. It's a tanh-knee waveshaper that's transparent below 0.7 of full scale and only bends peaks above that, so it catches the occasional cough or door slam without squashing the speech. It's a limiter, not a compressor, so there's no attack or release to set.
4. If it's loud but still hard to follow, load the Voice preset on the equalizer and tweak 230 Hz and 3 kHz as described above.
5. Export. Output is WAV at 16, 24 or 32-bit float, rendered offline faster than real time. No watermark, no sign-up, no length cap beyond what your device's memory can hold.

WAV is big: about 10 MB per minute at 16-bit stereo 44.1 kHz, roughly 5 MB per minute in mono, which is what most voice memos are. For a 40-minute interview that's a large file to move around. There's no MP3 export here, so if you need to email it, convert the WAV to M4A or MP3 in whatever converter you already use.

What this can't fix

• It can't turn your phone or laptop volume up. No web page can, and no iOS app can either — that's an OS-level control. It makes the file louder, which is the part that was actually the problem.
• It can't remove background noise, separate voices, or isolate one speaker. None of that is here.
• It can't recover clipped audio. If the memo is quiet but the loud moments already sound crunchy, that damage happened at record time and the flattened peaks are gone. Boosting gives you a louder clipped file.
• It can't touch DRM-protected audio, so anything playing inside a streaming app is out.
• It measures loudness, it doesn't silently re-master. The LUFS reading tells you how far you are from a target like -16 LUFS for Apple Podcasts or -14 LUFS for Spotify and YouTube. You read the number and move the slider, or click the suggestion. Nothing happens behind your back. If you're prepping a memo for a podcast feed, the guide to LUFS targets per platform lists what each one wants.

Recording so it doesn't happen again

Phones don't give you a manual record level, so the fix is physical. Get the microphone within arm's reach of whoever is talking. Sound pressure falls off fast with distance, and halving the distance is worth about 6 dB — more than any EQ move you'll make afterwards. Know where the mic is on your phone (bottom edge, beside the charging port on most iPhones) and don't cover it with your palm or lay the phone face-down on fabric. Take the case off if it wraps the bottom. For a lecture or a doctor's appointment, put the phone on the table between you rather than in a pocket.

If you record this kind of thing often and want the boost on the phone instead of in a browser, the iPhone app and the Android app carry the same five EQ bands and the same Voice preset values, keep a library of your files with settings saved per track, and work offline. They go to 200 % rather than 500 %, and they have no limiter and no level-matched A/B, so for a badly quiet memo the web tool does more. For a repeat workflow on files that are already close, the app is faster.

Two neighbouring situations get their own writeups: a quiet lecture recording, where distance is the whole story, and pushing volume without distortion, which goes deeper on the limiter and on what clipping actually sounds like.