“The sound is bad.” That is the sentence attached to almost every take we are sent to rescue. It describes a discomfort. It names no fault. And until the fault has a name, nobody can say whether the file takes an hour of studio time or a trip back to the microphone.
You think you have an audio quality problem. What you actually have is one of seven, and those seven do not get treated the same way. Some disappear in post with nobody ever guessing they existed. Two cannot be repaired at all, and those two happen to be the most common.
Listen in the right conditions first
Before you clean up a recording, you need to know what you are cleaning. And a fault will not let itself be caught on laptop speakers, still less in a car. Take closed-back headphones and a quiet place, and play the file twice: once following the meaning of the words, once ignoring it completely so you hear only the sound itself, the silences, the ends of sentences, whatever hangs in the air between words. The second pass is the one that gives you the diagnosis.
One aside: background noise, hiss, air conditioning, traffic, is a family of problems in its own right, with its own remedies. It sits outside the scope of this grid.
The voice sounds distant and hollow
What it sounds like: the voice seems picked up three steps from the mic when it should sit close to the ear. Turn the volume up and it does not come closer, it only gets louder. Listen to a silence between two sentences: the room keeps ringing for a fraction of a second after the last word.
Where it comes from: the room, nearly always. The mic picks up the direct voice and, right behind it, every reflection off bare walls, a hard floor, the desk, the window. The gear has nothing to do with it. The room was recorded at the same time as the voice, and the two are now mixed into one signal.
What can be done: de-reverb processing, which tries to separate the voice from the room around it, works while the reverb stays discreet. As soon as you can plainly hear it, the tool takes it out and carries part of the voice with it. The timbre thins, the consonants go dull, and the voice takes on that filtered, slightly underwater character an ear spots at once without being able to name it. You have traded one fault for another, often a less acceptable one.
Verdict: partly fixable while the reverb stays light, re-record as soon as you can hear it in the silences. It is the most common problem, and the one that resists the most.
The sound spits or tears on the loud words
What it sounds like: a short crackle, a dirty tear, a smeared sound, always on the loudest passages. The murmured words are spotless, the energetic ones are damaged.
Where it comes from: clipping at the recording stage, a signal that arrived louder than the chain could take. It is worth seeing what that means. The part that went over was not recorded too loud: it was not recorded at all. It is absent from the file.
What can be done: declipping tools estimate the missing shape and glue the pieces back together. On an isolated accident, one word that snaps in the middle of a clean take, it holds. On a take that clips from the first word to the last there is nothing to rebuild, only to guess, and the ear hears the guess. The objection always comes: “we’ll turn it down in the edit.” Turning down a clipped signal gives you a quieter clipped signal, nothing more.
Verdict: re-record. It is the only fault on this list that is strictly irreversible.
The p’s and b’s blow up the sound
What it sounds like: a dull thud, a gust of low end that shakes the voice on certain words. These are plosives, the consonants that push a puff of air straight onto the microphone capsule. The cause is nearly always the same: no pop filter between the mouth and the mic, a mouth too close, or too squarely on axis.
Good news: a plosive sits at the bottom of the spectrum for a very short moment, so it is treated locally, without touching the rest of the sentence. Only the violent kind is a real problem, the one that moves the voice itself under the blow: it leaves a dip in the word.
Verdict: fixable in the vast majority of cases.
The voice is too quiet, or the level jumps between sentences
Two different problems under the same feeling. A voice that is uniformly too quiet gets turned up, that is arithmetic, and the only question is what comes up with it. A level that jumps from one sentence to the next tells another story: the person moved during the take, leaned in on the passages they wanted to stress, pulled back to breathe.
Dynamic processing tightens the gaps, a few passages get ridden by hand, and the final pass lands on the target loudness of the channel you are delivering to, which differs between television, radio and the web. One nuance matters: if the timbre changes along with the level, volume is no longer the real subject. The person moved in front of the microphone, and you are on the next symptom.
Verdict: fixable.
The voice lacks body, or sounds thin and nasal
What it sounds like: a thin voice with no foundation in the low end, one that does not fill the space. Or the opposite, a voice bunched into the mids, with that telephone quality that tires an ear quickly.
Where it comes from: placement and mic choice. Too far from the microphone, off axis, aimed too high or too low against the mouth, or a mic whose character does not serve that particular voice. The low end of a voice is partly earned by proximity, and it drops away fast as soon as you step back.
What can be done: EQ, which balances the frequency bands of the voice, gives back a little foundation or deflates a band sitting too far forward. Improving a voice recording after the fact always means working with what the microphone kept: you sculpt what was captured, you do not manufacture what is missing. Pushed too far, EQ produces a voice that sounds worked on, and a listener feels that before understanding it.
Verdict: partly fixable. A small deviation gets corrected without anyone noticing. A voice captured plainly off axis will stay a voice captured off axis.
Mouth noise and saliva clicks
What it sounds like: small clicks between words, sticky mouth openings, mostly before sentences and after breaths. It comes down to a mix of physiology, hydration, vocal fatigue and proximity: a sensitive mic placed very close amplifies what an ear would never notice in conversation.
These noises come out one at a time. Automatic tools do the rough pass, the finishing is done by ear, sometimes click by click. The work is long and holds no particular difficulty. The trap is real though: erase too many and the voice starts to lisp while the consonants go blunt.
Verdict: fixable, at the price of the time you put into it.
Everything is clean, and it still does not hold
What it sounds like: nothing wrong, precisely. The sound is clean, the level is right, no noise sticks out. And you tune out by the third sentence. The pace stays flat, every sentence lands the same way, the words that carry the message go by unmarked. You understand all of it and remember none of it.
Where it comes from: no direction. The text was read correctly, without anyone deciding what it was supposed to do to a listener. Who it speaks to, what it has to trigger, where it picks up, where it settles, which words carry the intention. A text read and a text performed give you two files of equal technical cleanliness and two effects with nothing in common.
What can be done: nothing. No processing lays an intention in after the fact. You can tighten the edit, slide a tenser piece of music underneath to fake some energy. That moves the problem by a few seconds.
Verdict: re-record, with voice direction this time. No tool in the audio chain touches that fault, and it is the most common one on takes that are technically beyond reproach.
From the booth
People assume a higher-end microphone will settle the problem. In the booth the effect is often the reverse: the more sensitive a mic is, the more faithfully it renders the room it stands in, reflections included. A modest mic in a treated space gives you a usable take, while the same budget spent on the mic instead of the acoustics produces a very faithful capture of a bad room. A take done in a closet full of clothes sometimes sounds better than one done in a handsome glass-walled office.
The verdict at a glance
| What you hear | Likely cause | Verdict |
|---|---|---|
| Distant, hollow voice, silences that ring | Room reverb | Partly fixable, often a re-record |
| Crackle on the loud words | Clipping at the recording stage | Re-record |
| Thuds on p’s and b’s | Plosives, no pop filter | Fixable |
| Voice too quiet, or level jumping | Gain set wrong, unstable distance | Fixable |
| Voice with no body, or nasal | Placement and mic choice | Partly fixable |
| Clicks, mouth, saliva | Proximity, session hygiene | Fixable, at the price of time |
| Nothing wrong, but you tune out | No direction | Re-record, directed |
Two things jump out of the right-hand column. The most widespread faults, the room and clipping, are exactly the ones post-production does not save. Everything else gets corrected, in part or in full, provided you accept the time it takes and know when to stop, before the cure is more audible than the problem. Improving the audio quality of a recording always starts with that sorting.
The question that comes right after
A file repairs up to a point, and that point gets judged by listening. What is left is the thing nobody asks out loud. When the recording carries a brand’s name, when it is going to run for months on a campaign or inside a training module, the question is not how far it can be pulled back: it is whether the result will be worth what it stands for.
If you have a file in that state, have it listened to before you decide. We will tell you what repairs, what does not, and what a new take means for your schedule. Fifteen years of sessions makes that call fairly quickly, and our team works from its studio in Aix-en-Provence. To have a take listened to, or to scope a new recording, tell us about your project.
Not sure about a file?
Have the studio listen to it before you spend on repair work. We will tell you honestly what can be corrected and what needs a new take.
Contact usRead also
- How to record a voice over: the method, and what a studio changes: get a clean take from the start
- Voice over mistakes: what derails a project, and the warning signs to watch for: the scoping mistakes that produce these files