Screen recording audio too quiet — how to fix the recording you already made

Yes, you can fix it now, and you don't need to re-record. A screen recording that's too quiet is almost always just low in level, not damaged, and raising the level afterwards costs you nothing as long as you don't push it into clipping. Open the file in the browser volume booster, click Smart Boost, and it tells you exactly how much gain the file can take before a single sample clips. Apply it, export, done. Nothing gets uploaded, because there is no server.

That last part matters more than usual here. Screen recordings contain client calls, internal demos, unreleased product screens and whatever was sitting in your notification bar. Sending that to a "free online booster" that processes on someone else's machine is a poor trade for a gain change your own browser can do in a few seconds.

The five-minute fix

Do this first. Read the rest afterwards if you want to know why it happened.

1. Open the volume booster and drop the audio in. It loads whatever your browser can decode, so WAV, MP3, M4A/AAC and usually FLAC. If your recording is still an MP4 or MOV, see the next section.
2. Click Smart Boost. It scans every sample and reports sample peak in dBFS, RMS in dBFS, integrated loudness in LUFS, and your remaining headroom in dB. It also names the exact percentage that leaves zero clipped samples, keeping 1 dB of margin below full scale.
3. Apply that number. If Smart Boost says 380 %, your file was sitting about 11.6 dB below where it could be.
4. Leave the soft limiter on. It's a tanh-knee waveshaper: transparent below 0.7 of full scale, asymptotic to full scale above it. It catches a stray transient (a keyboard clack, a chair squeak) without flattening the speech around it. It's a limiter, not a compressor, so there's no attack or release to set.
5. Export as WAV. 24-bit is the sane default; 16-bit and 32-bit float are there too. Rendering happens offline, faster than real time.

One honest technical note: Smart Boost measures sample peak, not true peak (dBTP). In a lossy-encoded source, the reconstructed waveform between samples can sit a fraction of a dB above the highest sample value. The 1 dB of built-in margin covers that in normal use, but if you're delivering to a spec that demands −1 dBTP, check it in a meter that does true peak.

If the recording is still a video

Screen recordings usually arrive as MP4 or MOV, and that changes the workflow slightly.

For most people that's fine, because a quiet screen recording is usually going into an editor anyway to have the dead air trimmed. If you genuinely just want the same file, louder, with no editing, the shortest route is to pull the audio out with your editor, fix the level here, and export once at the end.

Why screen recordings come out so quiet

It's rarely one cause. Usually two or three stack up.

The OS captures the mic conservatively. macOS and Windows both set input gain low by default and neither applies makeup gain afterwards. A USB condenser 40 cm from your face at a default input level often lands around −30 dBFS peak on speech. Technically that's fine. It's clean, just small. It's also why QuickTime screen recordings sound thin even when the mic is good.

System-audio capture inherits whatever was playing. If the call you were recording was at 40 % system volume, the capture is 40 % of what it could have been. Nothing normalises it on the way in.

Mixing mic and system audio drags both down. Two sources summed together can clip where either alone wouldn't, so plenty of capture setups pull both faders back to stay safe. Each source then lands several dB lower than it would have on its own. Two quiet sources plus a safety margin is how you end up with a 40-minute file peaking at −28 dBFS.

OBS defaults aren't loud. The mic slider starts at 0 dB with no Gain filter attached, and the meters are peak meters, so a source that looks like it's moving can still average −35 dBFS. The usual advice for "OBS recording too quiet" is to add a Gain filter and record again. If the recording already happened, you don't have to.

How much boost, in decibels

Percentages are easier to click, but decibels are what your ears track. Doubling amplitude is +6.02 dB, so 200 % is +6 dB and 400 % is +12 dB. Here's the mapping:

There's a target-based route as well. Integrated loudness is computed with a real ITU-R BS.1770-4 implementation: K-weighting filters, 400 ms blocks with 75 % overlap, an absolute gate at −70 LUFS and a relative gate at −10 LU. So if your file measures −29 LUFS and you're posting to YouTube, which plays back around −14 LUFS, you know you're roughly 15 dB short. Apple Podcasts sits at −16 LUFS. Worth knowing: YouTube and Spotify turn loud uploads down, they don't push quiet ones up, so a quiet upload stays quiet forever. The tool shows you the measured number and the suggested percentage. It doesn't silently re-master the file to a target behind your back. You read it and decide, or click the suggestion.

A live clip warning sits under the slider too, naming what percentage of the file would run past full scale at your current setting. If that number is anything above zero and the limiter is off, back the slider down.

Speech needs EQ, not just brute gain

Screen recordings are almost always talking. Voice energy sits mostly between 100 Hz and 8 kHz, and intelligibility lives around 2–5 kHz, where the consonants are. Room rumble, laptop fan and desk thumps live below about 200 Hz and carry no information at all. Push a quiet voice recording up 12 dB flat and the fan comes up 12 dB with it.

Try the Voice preset in the equalizer before reaching for more gain. It's five peaking biquads at 60, 230, 910, 3000 and 14000 Hz with a Q of 1.0, and Voice sets them to −4, −2, +3, +5, +2 dB. The cuts at 60 and 230 Hz take out rumble and boxy chest thickness; the +5 dB at 3 kHz is the one that makes words land. Because you've removed low-frequency energy that was eating your peaks, the file will usually take another 2–4 dB of level afterwards. So run Smart Boost after the EQ, not before.

The level-matched A/B button is what tells you whether you actually improved anything. One click switches between processed and untouched, and the processed side is trimmed back to the original's peak automatically so both sides sit at the same loudness. Louder always sounds better for the first two seconds, which is how people talk themselves into bad EQ. Matched, you're comparing tone instead. If the Voice preset sounds worse at matched level, it is worse. No other free tool does this.

Skip the Bass Boost preset here (+10, +5, −2, 0, +1). It's built for music. On a narrated demo it adds mud and eats the headroom you were trying to gain. Same story with the separate low-shelf bass tool at 120 Hz: good on music, wrong direction for speech.

What this can't fix

Being straight about the edges saves you an hour of trying.

• Distortion already in the file. If the original clipped while recording, the tops of the waveform are already flat and that information is gone. Gain doesn't rebuild it; it just makes the crunch louder.
• Background noise. There's no noise removal here. Gain lifts hiss and voice by the same amount, so a file with a −45 dBFS noise floor boosted 15 dB has a −30 dBFS noise floor and exactly the same signal-to-noise ratio. EQ helps a little by cutting the rumble band. It isn't denoising.
• Splitting your voice from the system audio. Once they're mixed into one track they're one track. There's no stem separation.
• System or device volume. Nothing in a browser can push your phone or laptop output past its maximum, and no app can either, because that ceiling is enforced below the app layer. What these tools raise is the level inside the file, which is the thing that was actually wrong.
• DRM-protected audio can't be loaded or processed.
• MP3 export. WAV only, at 16-bit, 24-bit or 32-bit float. Editors and podcast hosts prefer WAV anyway; if you need MP3 for delivery, encode it at the last step.

So it doesn't happen again

Aim for speech peaks around −12 to −6 dBFS while recording and none of this comes up. In OBS, add a Gain filter to the mic source and set it during a rehearsal take so peaks sit in that band, then watch the mic and desktop-audio faders separately instead of trusting the mixed meter. In QuickTime, choose your actual interface or USB mic from the dropdown next to the record button rather than the built-in default, and raise the level in System Settings, Sound, Input until the meter moves properly on normal speech. On Windows, the Levels tab often has two controls: a percentage slider and a Microphone Boost field in fixed dB steps, and the boost one is usually the one still sitting at 0 dB.

Record twenty seconds and listen back before the real take. It costs far less than repairing an hour-long file. And when the file you need to rescue is on your phone rather than a laptop, the Increase Volume app for iPhone has the same five EQ bands and the same preset values with per-file settings saved in a library, and there's an Android version too. The limiter and the level-matched A/B are website-only for now.