Fix muffled audio
Voice sounds dull, like it's behind a door? fix.fm checks why it's muffled and brings back the clarity. Free, in your browser, nothing uploaded.
Never uploaded. Everything runs in your browser.
What's happening
Muffled audio sounds dull, like the person is talking through a door or with a hand over their mouth. The words are there, but the crisp top of the voice, the part that makes s, t and f sounds clear, is missing. It usually comes from something in the way: a lav mic under a shirt, a face mask, a mic pointed away from the mouth, or a headset that cuts the highs.
fix.fm compares the top of the voice, from 4 to 10 kHz, with the middle where most of speech sits, after taking out any hiss so it can't pass for clarity. Once the top is more than 34 dB under the middle you start to hear the muffle, and past 50 dB it sounds like it's coming from the next room.
How far the EQ goes
EQ can bring back detail that was recorded quietly. It cannot bring back detail that was never recorded, and it cannot turn hiss into speech. The three cases below are the same clean reading with its top end taken away in software, so we know exactly what was lost and can measure how much came back.
One clean reading, damaged three ways in software, then run through this page's fixer at full amount. Every clip is matched to -23 LUFS so only the tone differs.
Clean reading
The reference. Listen to how s, t and f sit on top of the words.
Low-passed at 3 kHz
A hand over the mic, a lav under a shirt. The words are all there and the consonants are gone.
3 kHz case, after the fix
The consonants are back. Compare with the clean reading: close, with a slight hardness on s.
Low-passed at 1.5 kHz
Through a wall. Almost nothing above the vowels survived.
1.5 kHz case, after the fix
Brighter and easier to follow, but still dull. There was too little left to lift.
3 kHz with hiss
The same dull voice with steady hiss 30 dB under it. The hiss sits exactly where the lost highs were.
Hiss case, after the fix
The hiss came out first, then the lift. Without that step the hiss would have come up with the voice.
Saved at 8 kHz
Phone-call quality. The check names it as the format, not the voice, and the fixer leaves it alone.
Public-domain LibriVox voice. The damage is a software low-pass, added hiss and a sample-rate change, not a real covered microphone or a real call, so the clean version really is what was lost.
| Case | Highs before | Highs after | Hiss before | Hiss after | True peak |
|---|---|---|---|---|---|
| Dull, 3 kHz | -54.7 dB, severe | -20.9 dB, fine | 66.1 dB SNR | 63 dB SNR | -5.8 to -2.8 dBTP |
| Dark, 1.5 kHz | -78.7 dB, severe | -62.7 dB, severe | 65.3 dB SNR | 66 dB SNR | -8.5 to -6 dBTP |
| Dull with hiss | -48.8 dB, moderate | -31.3 dB, fine | 33.6 dB SNR | 65.3 dB SNR | -5.8 to -2.5 dBTP |
Read the improvement and the remaining loss separately. The 3 kHz case ends within a couple of dB of the clean reading. The 1.5 kHz case gains 16 dB and is still graded severe, because the boost stops where the voice runs out: no band is lifted further than the voice stands above the noise there. That cap is what keeps a dark recording from turning into a bright hiss, and it is also why the fix cannot promise to clear every severe file. For a take like that, the usable move is to accept a duller voice and make sure the noise and level are right. How we test has the pass marks behind these grades.
How it works here
- 01
Measures what's missing
It compares the crisp top of your voice with the middle, and works out how much of the top was lost.
- 02
Brings the top back
A rising EQ shaped to the loss lifts the highs, only as far as the voice stands above any hiss. If hiss is holding it back, the hiss comes out first, and the result stays under -1 dBTP.
- 03
Stays on your computer
Your file is never uploaded. The EQ runs in this browser tab.
Questions
Can it fix audio recorded through a mask or under a shirt?
Usually, yes. Those take the top off the voice but leave it underneath, and the EQ can lift it back up. In the examples above, a voice cut at 3 kHz comes back to within a couple of dB of the clean reading. The more that was blocked, the less there is to lift, so very muffled audio gets clearer but not clear.
Why does the fixed file sound brighter but still dull?
Because the boost stops where the voice runs out. Below about 1.5 kHz of cutoff, what sits above the voice is mostly noise, and lifting it would add hiss, not words. The fix lifts what is there and leaves the rest, and the grade afterwards tells you honestly how much is left.
Is my audio uploaded anywhere?
No. Everything happens inside this browser tab. The file is never sent to us or anyone else.
Why does the fixed version sound a bit hissy?
Lifting the highs also lifts anything else up there. fix.fm limits the boost to stay above the noise, but if the recording has hiss, pick the noise fix too. It runs first, so the EQ has less to bring up.
Why does my phone call recording sound muffled?
Many calls are sent at 8 kHz, which cuts everything above 4 kHz before it reaches you. That sound is gone, not hidden, so there's nothing to bring back. fix.fm tells you when a file is like this instead of making it worse.
How do I avoid muffled audio next time?
Point the mic at your mouth, keep clothing and hair off a clip-on mic, and don't wear a mask while recording if you can help it.