How to improve your voice recording quality (2026 refresh)
Updated: Oct 6
If you're thinking of doing any voice over work from home, there's a few things you can do to make your audio sound better. If you haven't yet treated your space or don't need a pro set up (if you're podcasting for fun, for example), you can still enhance the quality with some easy to implement tricks. Well, they're not tricks really, but I couldn't think of how to end that sentence.
I first wrote this post back in 2021, and a fair bit has changed since then. And by 'fair bit', I mean loads. The tools at our disposal are cleverer, AI has gatecrashed the cool kids party, and I've learned a whole bunch of new stuff along the way. So the blog has had a proper refresh.
ONE: Stop the noise before you start (and please go easy on the clean-up).
If you have background noise in your audio, you need to take a step backwards. And by that I mean get rid of it BEFORE you start.
Close windows, shut doors, turn your phone off, throw the fridge away, tell people to shut up.
Get under a duvet, build a pillow fort, line a cupboard with mattresses.
Whatever you do, it’s easier to stop background noise at source than it is to fix it later.
"But Martin, there's software that removes noise now! Facebook ads keep telling me about it ALL THE TIME. " Ahem.
Well, yes, there is.
iZotope RX is one of the most popular and well known on the market. Its de-noising tools can pull out hum, hiss, air con and computer fans surprisingly well. (Trust me, I have a client who records in a room with servers buzzing away at a mad volume and how RX can deal with it is pretty ruddy great.)
There are free options too, hurrah, like those built into software you may already have. Audacity has the Noise Reduction effect, and then there are one-click AI "enhancers" such as Adobe Podcast Enhance.
BUT, and it's a big but (with capitals), please go easy.
Unless you're assisting by telling the software what to remove, it's likely you're dragging down a slider somewhere to reduce the noise. Push it too hard and it starts going wrong. That's when you can get:
A watery, "underwater" or slightly robotic sound (these are called artefacts)
A thin, lifeless voice, because the warmth and air get stripped out along with the noise
Swirly, warbly sounds in the gaps between words
Breaths and word endings that get chopped, so the room seems to vanish and reappear between phrases
I've been in videos (on camera) where the producer has been so heavy with noise removal that things like footsteps, hand claps, leg slaps etc all get lost and it ends up sounding and looking weird.
Over-processed audio will suggest "my room isn't up to scratch" or "I'm hiding something". Most clients would rather have natural audio with a quiet, consistent room tone than something that's been blasted with dampening.
My rules of thumb:
Use the lightest setting that does the job, then back it off a notch
Do two or three gentle passes rather than one heavy one
Flick the effect on and off and listen on headphones
Always keep your raw, untouched file.
Think of these tools as a safety net rather than the rule.
TWO: Record it properly in the first place!
A lot of "bad audio" is really just badly recorded audio. If you get the basics right then everything after is much easier. Here's some free scripts to test your setup.
Set your levels. Aim for your loudest bits to peak somewhere around -12dB to -6dB. Too quiet and you'll raise the noise when you boost it later. Too loud and the audio will clip, which is distortion you can't fix (the peaks go past the point of no return).
Record at 24-bit, 48kHz. It's a widely used standard and gives you plenty of headroom.
Stay put. Keep a consistent distance away from the mic, roughly a hand-span for most people. (That hand span is the "shaka" sign). Wander about and your volume and tone wander with you. But also, if you have a quiet bit, you can lean in... and a louder bit, hang back.
Tame the pops. Use a pop filter, and angle yourself slightly off the mic rather than talking straight into it. Your Ps and Bs will thank you. Actually, everyone listening will thank you, they just won't realise why.
Grab some room tone. Record five to ten seconds of silence at the start. It's handy for noise reduction later, as the tools work better with a good sample of the room to work from.
THREE: EQ and the high pass filter.
After recording voice, one of the first things you should do is EQ it. Something I apply to all my VO work IMMEDIATELY is called a high pass filter. (Actually, these days I have VO templates set up so it's already added before I record).
As the name suggests, it allows high frequencies to pass through but cuts off lower frequencies.
Make a cut around 80Hz to purge low frequencies - those which are lower than the human voice. You can do this in your audio editing software.
This picture shows exactly that, in Cubase.
Don't get carried away and push it too much, though. Go too far and voices, especially deeper ones, start sounding thin and tinny. Bad. Unless you're from The Wizard of Oz. (It's a tin man joke).
If your recording sounds a bit boxy, a small cut somewhere in the 200 to 500Hz range can help.
Make small moves of a couple of dB and compare before and after. With EQ, less is almost always more. And think of cutting before boosting.
FOUR: Compression.
Voices are extremely dynamic so you're likely to have a file that has some loud parts and some quiet parts. To stop those parts being so obviously different, you need to compress it.
No, not with Winzip. You won’t be able to hear it at all then. (Like I said I wrote this blog a while ago, but I'm keeping that old joke).
You’ll find the compressor in your audio editing software.
A compressor brings the louder parts closer to the quiet parts by squishing them down, giving you a more balanced waveform.
Think of it LIKE THIS text you're reading, and every now and THEN, a word comes out louder THAN the OTHERS. Insert a compressor and it makes those capitals into the same case as the rest of the sentence. But in audio form. Yes, that was a weird way to describe it, I agree.
To try an explain it a bit more clearly, these two images show the original voice recording and then the compressed voice recording. You can clearly see where the peaks have been tamed.
The same warning as the noise reduction comes in here: don't overdo it. Squash too hard and your voice loses its life. Breaths and background noise get louder too, because everything is being pushed towards the same level. A gentle ratio of around 2:1 to 4:1, taking a few dB off your loudest words, is a sensible starting point.
FIVE: De-essing and de-clicking.
Compression can come with a side effect.
It can make your S sounds sharper and your mouth noises (GROSS) more obvious.
Sibilance is that harsh hiss on S and SH sounds. A de-esser turns down just those frequencies, only when they happen. Use it lightly, or you'll end up sounding like you've lost your front teeth.
Mouth clicks are the little lip smacks and sticky noises from a dry mouth. When you hear them, you can't stop hearing them. It's a curse. RX has a Mouth De-click tool that's great for these, or you can edit the worst ones out by hand (easier than you might think). Better still, prevent them. Sip water, avoid dairy before a session, and try a slice of green apple.
I once had to edit an audiobook for someone. Only a short one, so I thought maybe after I would do the clicks by hand. But after the session I ran RX across it and it found over 3,000 clicks. HYDRATE. HYDRATE. HYDRATE.
SIX: Normalising and loudness.
Right. So you think the file is too quiet after compressing it? What you need to do now is normalize it to -1dB.
Again, this is something you'll find in your audio editing software.
Normalizing audio boosts the lovely even sounding file from point FOUR to a decent volume so you don’t have to strain to listen to it.
Here's the compressed waveform again and below it is the normalized version. It's the same audio, just louder.
An important update since I first wrote this. Normalising to a peak level is a quick fix, but it doesn't tell you how loud something actually sounds. These days most platforms measure loudness in LUFS.
Around -16 LUFS is a common target for podcasts.
Audiobooks for ACX have their own specs, with average levels between -23dB and -18dB and peaks no higher than -3dB. Most editing software now has a loudness meter, or a loudness normalise option.
Quick note: There's some jobs I normalize, some I don't. When I don't normalize, I'm usualy compressing/limiting instead. If in doubt, do neither. And if a client gives you technical specs, follow them.
SEVEN: Gates (and their gentler cousin).
After normalizing you might start to hear room noise and other annoying things in the background.
You should have read point ONE. Tut, tut.....
....or you can apply something called a gate. You'll find this in your audio editing software.
A gate stops any sound below a certain dB level coming through. Cars driving past, someone talking in the next room, a fart (depending on the intensity).
The first image below is some room noise/ambience, which is quiet but will intrude and affect the quality of the recording. The second picture shows the effect of applying the gate. No more room noise. MAGIC.
It's worth noting, if your room is so loud you can hear it at the same time as the voice in your recording, a gate won't get rid of it.
Two more things worth knowing. First, a gate set too tightly can clip the starts and ends of words, so give it a slightly longer release. Second, total digital silence between phrases can sound oddly unnatural. An expander is a softer alternative. Rather than shutting the noise off completely, it just turns it down, which sounds far more natural. And that's what we all want.
EIGHT: Editing voiceover by hand.
If a gate or expander is too heavy handed or scary for you (it can be tricky getting the balance right), this is how to edit a voice recording.
Here is a short waveform of a voice.
This is the same clip but the ends have been trimmed. Depending on your software, it's usually a case of just dragging the ends in which removes the audio before and after the voice.
This is the same clip but the beginning has a fade in and the end has a fade out. If you don't fade in/out, you leave yourself open to a click or pop noise as the clip ends. Adding a short fade ensures smooth start and end points.
This is the same clip but the noise/breath between the two parts has been removed. You cut it out, then add the fades as above. You can do this to get rid of coughs, mistakes, loud breaths, or anything else you don't want in the audio. If you record podcasts, you'll probably be doing this to cut down on your episode length, taking out long pauses, umms, ahhs, rambling speeches that add no value whatsoever.
A word of caution: don't remove every single breath. People breathe, and a read with no breaths at all sounds robotic. Turn the loud ones down instead of cutting them, and keep the natural rhythm of the read. Having said that, more often than not, commercials have all the breaths removed, unless it's a part of characterisation in the script. Commercials often have too many words stuffed in so there's no room for breaths. And some people see the breathing as distracting from the message the ad is trying to convey.
NINE: Listen like your audience does.
You've spent an hour staring at waveforms and listening on your best headphones but your listener is on a phone speaker in a noisy kitchen.
Before you call it finished, check your audio on whatever you have available:
Good headphones
Cheap earbuds
Your phone speaker
The car
Then take a break and come back with fresh ears - this is VERY important. Not just because your ears get tired, but you can also just lose focus. When you come back to it, problems you'd stopped noticing will jump right out. It also helps to compare against a professional recording you like, so you have a benchmark.
TEN: Know when to leave your audio alone.
Plenty of clients and studios don't want you to process audio at all. They want it "dry", meaning clean and unprocessed, so their own engineer can mix it to match everything else. They often have their own signature sound.
So before you reach for the (admittedly very cool) plug-ins, check what the client actually wants. Deliver the format they ask for (WAV or MP3, plus the sample rate), name your files clearly, and keep your raw recordings safe in case they need a re-edit.
Then try your hardest to forget perfection. The more you fiddle, the further you'll move away from what you actually want.
THE END
There you have it, how to improve voice recording quality. This is by no means a comprehensive list and there are many other ways to tidy things up, but the points here should give you a nice start, rather than just having to use the audio straight out of the box.
If there's one thing to take away from this excessively long post, it's this... a quiet room and good technique beat clever plug-ins every time. Process gently, keep your raw files, and trust your ears (after a cup of tea and a break).
If you'd like to talk to me about cleaning up audio or voice work, please do get in touch via the contact page. And if editing isn't your idea of fun, my voiceover demo production service includes all the mixing and mastering.
