Free tools / Vocal compressor
even it out, warm it up
Vocal compressor
Drop in a take where some words jump out and others get lost. It brings the quiet ones up and settles the loud ones, adds a little warmth if you want it, and saves a new WAV. The settings are deliberately gentle, so it still sounds like you.
Your file is read and processed on your device. It isn't uploaded anywhere.
How much
Start with one of these
Fine-tune it
Slowly brings quiet sentences up and loud ones down towards your average level.
How firmly the louder words are held back. 1:1 is off; it stops at 6:1.
Relative to the average level of your voice in this take, so it behaves the same on quiet and loud recordings.
A gentle lift below about 180 Hz for a little more body. It stops at +4 dB.
A small lift around 4.5 kHz to help words cut through. It stops at +2.5 dB.
Level and file
Use the loudness the brief asks for. The loudness normaliser explains the common targets. A peak limiter keeps the loudest moments under the ceiling.
your take
Choose a file to start.
Before
After
What changed, moment by moment
turned upturned downcompared with the take's typical level, before the final loudness change
Switch while it plays to hear the difference at the same moment. Headphones help.
Before
- Loudness
- Loudness range
- Word-to-word spread
- True peak
After
- Loudness
- Loudness range
- Word-to-word spread
- True peak
What the vocal compressor does
It works in a few gentle steps, in this order:
- Rumble filter. Takes out very low sound below 60, 80 or 100 Hz, like traffic or a desk being knocked, which a voice doesn't need.
- Evening out the phrases. A slow leveller looks at each part of the take over about half a second and moves it part of the way towards your average level. Quiet sentences come up and loud ones come down. It holds its setting through pauses, so it doesn't lift the silence between lines.
- Compression. A faster stage holds back the louder words and syllables, so the voice sits at a steadier level. It measures your voice's average level first and works relative to that, so the presets behave the same on a quiet take and a loud one.
- Warmth and presence. Optional, small tone changes: a lift below about 180 Hz for a little more body, and a lift around 4.5 kHz to help words cut through.
- Softening harsh S sounds. Optional. When an S or SH sticks out, it turns down only the top of the sound, above about 6 kHz, for that moment.
- Level. Finally it sets the loudness you chose, measured in LUFS, and a look-ahead peak limiter keeps the loudest moments under your true peak ceiling.
Everything happens in your browser. Your file isn't sent anywhere, and nothing is saved unless you press Save.
What the numbers mean
- Loudness range is the spread between the quieter and louder parts of the take, measured over three-second windows as described in EBU Tech 3342. A smaller number means a more even take. It needs a take of a few seconds or more.
- Word-to-word spread is the same idea over shorter, 0.4-second windows, so it shows how much individual words jump around. It's our own measure: the gap between the 95th and 10th percentile of momentary loudness.
- True peak is the highest point the waveform reaches, including between samples.
Getting a natural result
- Clean up and record quietly first. Compression brings the quiet parts up, and that includes room noise, breaths and mouth noise. Record in the quietest spot you can (the room noise check helps), and try the gentle clean-up before this tool if there's hum or a steady background noise.
- It can't fix echo. A boomy or hollow room still sounds boomy, and compression can make the echo more noticeable.
- Start light. Listen with the loudness matched, so the processed version doesn't sound better just because it's louder. If the voice starts to sound flat or squashed, go back a step.
- Keep it sounding like you. Voicefolk's rules ask voices to keep their audio natural, with no heavy processing, so the presets are deliberately gentle and the controls stop before extreme settings. Nothing here changes your pitch, and the tone controls stop at small amounts.
Before you send it
Some clients want processed audio and some want it "dry", with no compression or EQ, so they can process it themselves. Ask before you send, and keep your original file either way. If they want dry audio, the loudness normaliser changes only the volume.
More free tools: record and check
Room noise check
Measure how quiet a room is with your own mic, and compare spots at home to find the best place to record.
open itAudio file checker
Drop in a take to see its sample rate, bit depth, peaks, clipping and room tone, and check it against a spec.
open itLoudness normaliser
Measure a take's loudness in LUFS and true peak, and save a copy at the level the client asked for.
open itGentle clean-up
Take out low rumble and mains hum, and lightly lower steady background noise, without changing your voice.
open itAutocue and recorder
Read your script from an autocue and record audio or video on the same phone.
open itAnd the rest
got a voice for it?
Professional voiceover artists can register for a lifetime listing, open to the UK first (other countries coming soon). We'll be in touch as soon as the deal goes live.
Voicefolk is a directory of real voiceover artists, for brands who'd rather hire a person than generate one. No AI clones, and you agree your deals directly. Join as a voice