Fix audio that's too quiet
Hard to hear even at full volume? fix.fm measures how quiet it is and brings it up to podcast level without clipping. Free, nothing uploaded.
Never uploaded. Everything runs in your browser.
What's happening
Audio that's too quiet makes people reach for the volume and still strain to hear. It often happens when the mic gain was set low to be safe, the speaker sat far from the mic, or a call app recorded at a low level. The voice can be perfectly clear and still be hard to follow on a phone speaker or in a car.
fix.fm measures programme loudness in LUFS. Our check flags quiet speech below -27 LUFS, with stronger warnings below -31 and -38. The full-strength fix aims for -16 LUFS with peak control; your destination may require a different target. Compression and normalization explain why raising a whole file cannot balance every sentence.
How far the level fix goes
Turning a recording up is easy until something in it hits the ceiling. The level fix measures how loud the voice sounds, eases the loudest moments down, then turns everything up to -16 LUFS with the peaks held under -1 dBTP. The three cases below are the same reading made quiet three ways, so we can show what comes up with the voice, and where a plain volume change stops and this fix does not.
One reading turned down, then given hiss, then given one loud bang, each through this page's fixer at full amount. These clips are NOT level matched: the point is the level, so the quiet ones are quiet. Set your volume on a fixed clip first.
Quiet take, -36 LUFS
A clean reading recorded far too low. You will reach for the volume.
Quiet take, after the fix
Up at -16 LUFS with the peaks held under -1 dBTP. Put your volume back where it was.
Quiet take with hiss 30 dB under
The same take with a soft, steady hiss you may not notice at this level.
Hissy take, after the fix
The voice is up, and so is the hiss, by exactly the same amount. Listen to the pauses. This is why the noise fix runs first when you pick both.
Quiet take with one 60 ms bang
The same quiet take with a short burst in the pause at 4.5 s that already touches the ceiling.
Peaky take, volume slider
Turned up as far as a plain volume change can go before the bang clips: less than half a decibel. The voice is still quiet.
Peaky take, after the fix
The compressor and limiter hold the bang down, so the voice can come up to the target around it.
Public-domain LibriVox voice. The hiss is white noise and the bang is a windowed burst of noise, both made in software; nothing here was recorded quiet, noisy or with a real knock.
| Case | Loudness | True peak | Pauses | Loudest 400 ms |
|---|---|---|---|---|
| Quiet take | -36 to -16.5 LUFS | -15.5 to -1.2 dBTP | -89.6 to -67.1 dBFS | +2.8 to +1.4 LU |
| Quiet take with hiss | -36 to -16.5 LUFS | -15.5 to -1.2 dBTP | -64.9 to -42.4 dBFS | +2.8 to +1.4 LU |
| Quiet take with one bang | -29.8 to -16.9 LUFS | -1.5 to -1.2 dBTP | -89.2 to -69.1 dBFS | +7.5 to +1.6 LU |
| Peaky take, volume slider | -29.8 to -29.2 LUFS | -1.5 to -1 dBTP | -89.2 to -88.7 dBFS | +7.5 to +7.5 LU |
| Fixed take, through the fix again | -16.5 to -16.2 LUFS | -1.2 to -1.2 dBTP | -67.1 to -65.9 dBFS | +1.4 to +1.2 LU |
The quiet take is the plain case. It reads -36 LUFS, moderate, and comes out at -16.5 LUFS with the true peak at -1.2 dBTP. The loudest moment went from 2.8 LU above the average to 1.4: that is the compressor easing the strongest syllables so the rest can rise without the limiter working on every word. Run the fixed file through again and it moves 0.3 dB, the gap left to the target, and otherwise stays put; a file that is already loud enough is not made louder.
The hissy take is the honest part. The voice came up 19.5 dB and the pauses came up 22.5 dB, a little more than the voice, because the compressor only eases the loud syllables while the make-up gain is applied to everything in the file. The hiss was easy to miss at -64.9 dBFS and is plain at -42.4. That is why, when you pick noise and level together, the noise fix runs first.
The peaky take is why this is not a volume slider. One short bang already sits at -1.5 dBTP, so a plain gain change can add 0.6 dB before it clips and the voice stays at -29.2 LUFS, still mild. The fixer holds the bang down and brings the voice to -16.9 LUFS. The bang is still there, just no louder than a loud word; this fix does not remove bangs, clicks or plosives.
What this does not prove: these are one clean voice with a steady level. A recording where one speaker is far quieter than another is a different problem, because one gain for the whole file cannot close that gap; uneven podcast volume measures that case, and the loudness meter shows where a file is quiet before you fix it. How we test has the pass marks behind the check.
How it works here
- 01
Measures the loudness
It measures how loud your voice sounds to a listener, the same way podcast apps and streaming services do.
- 02
Turns it up safely
It brings speech up toward -16 LUFS, the usual level for podcasts and video, with a gentle compressor and a limiter so nothing clips. AMOUNT sets how far.
- 03
Stays on your computer
Your file is never uploaded. It all runs in this browser tab.
Questions
Why not just turn the volume up in my editor?
You can, until the loudest word hits the ceiling and clips. fix.fm evens out the peaks first, so the rest of the voice can come up further without distorting.
Is my audio uploaded anywhere?
No. The file is measured and fixed inside this browser tab and never sent anywhere.
How loud should a podcast or video be?
Around -16 LUFS is the common target for podcasts and online video, with peaks no higher than -1 dB. That's what fix.fm aims for. Spotify, Apple Podcasts and YouTube all play back near there.
Will it make the background noise louder too?
Yes, turning a recording up turns up everything in it. If there's a hiss or hum you can hear once it's louder, pick those fixes too. They run first, before the level goes up.
Can it make audio quieter?
No. This fix only turns quiet audio up. A file that's already loud enough is left alone.