How to Improve a Bad Recording
To improve a bad recording, first remove background noise with Vocal Cut's free Noise Reducer, then shape the tone with EQ and set the loudness last. Before any of that, listen and work out what's actually wrong, because hiss, echo, clipping and a quiet level each need a different fix, and some of them can't be fully fixed at all. Every tool here runs in your browser, so your audio never leaves your device.
Start by naming the problem, then find its tool and detailed guide in the table below, and apply the fixes in an order where each one helps the next.
Step 1: Diagnose what's wrong
Put on headphones (laptop speakers hide hiss and rumble) and play the first minute. Then skip to a loud passage and a quiet pause, where most problems show up most clearly.
If you can't name what you hear, open the file in the Spectrogram, a picture of a sound's frequencies over time. Time runs left to right and frequency bottom to top, so steady noise shows as a band that stays through the pauses: hum as thin lines near the bottom, hiss as a haze across the top. Speech comes and goes. How to Read a Spectrogram shows what each pattern looks like.
| Symptom | What it sounds like | Fix | Tool | Deep dive |
|---|---|---|---|---|
| Steady hiss, fan or AC | A constant wash or "shhh" under everything, loudest in the pauses | Noise reduction (Voice mode for speech, Music mode for music) | Noise Reducer | How to Remove Hiss from Audio |
| Hum or buzz | A low, steady drone, often with a buzzy edge | Music mode with the 50/60 Hz hum option | Noise Reducer | How to Remove Hum and Buzz from Audio |
| Wind | Low thuds and rumble that come in gusts | Noise reduction, then cut the lowest EQ band | Noise Reducer, Equalizer | How to Reduce Wind Noise in Recordings |
| Echo and room reverb | Hollow, distant, like a bathroom or empty hall | Voice mode, kept gentle; expect a partial result | Noise Reducer | How to Remove Echo and Reverb from Audio |
| Harsh sibilance | "S" and "sh" sounds that hiss or spit | A small cut at 6 kHz or the 12 kHz shelf | Equalizer | How to De-Ess Vocals With an Audio Equalizer |
| Muffled or thin tone | Dull, like through a wall, or small and tinny | Small lifts at 2.2 kHz (clarity) or 240 Hz (body) | Equalizer | How to EQ a Podcast Voice for Clear Speech |
| Too quiet or uneven | You keep reaching for the volume; some words vanish | Compression for evenness, then loudness normalization | Compressor, Volume Normalizer | How to Make Quiet Audio Louder (No Clipping) |
| Clipping or distortion | Crackle, fuzz or a flattened, harsh sound on the loudest parts | Can be softened, not restored | Equalizer, Audio Cutter | What Is Audio Clipping (and How to Avoid It)? |
Most bad recordings have two or three of these at once. Note them all, then fix them in the order below, not the order you noticed them.
Step 2: Apply the fixes in this order
The order matters as much as the settings, because each step changes what the next one works with.
1. Trim first
Open the Audio Cutter and remove the dead air, the mic bumps and the fumbling at the start and end. Use the Trim Start point and Trim End point sliders, check with Play Selection, then press Cut & Download. A loud bang left in limits how far the loudness tools can raise everything else.
The Noise Reducer accepts files up to 100 MB and 12 minutes long, so split anything longer into pieces at this stage.
2. Remove background noise from the audio
Drop the trimmed file into the Noise Reducer and pick a mode:
- Voice uses an AI speech model (GTCRN) built for podcasts, voice memos and interviews. It keeps the voice and removes what isn't a voice, so don't run songs through it. The first run downloads the small model to your browser, about 0.5 MB, and caches it.
- Music uses spectral processing, not AI, for hiss, hum and broadband noise on mixes and instruments. This is also where you switch on hum removal.
Press Reduce noise (Voice mode) or Reduce noise (Music mode), then use the slider between Original and Denoised to choose how much to keep. It starts at 75% and applies live; if the voice sounds watery, pull it back. How to Remove Background Noise from Audio covers both modes and the common fixes in more detail.
3. EQ, including de-essing
Take the denoised file into the Equalizer, a six-band graphic EQ with bands at 80 Hz, 240 Hz, 750 Hz, 2.2 kHz, 6 kHz and 12 kHz, each adjustable from -12 to +12 dB. Typical moves for a voice:
- Pull 80 Hz down to cut rumble, handling noise and leftover wind.
- Lift 2.2 kHz by a couple of dB if the voice is muffled.
- Lift 240 Hz gently if it's thin.
- Cut 6 kHz (or 12 kHz) by a few dB if the esses are harsh.
EQ comes after noise reduction because denoising changes the tone of the voice, so you want to judge EQ on the cleaned result. Any boost also raises the noise in that band, which would give the noise reducer a harder job.
4. Dynamics, if the level jumps around
If some words are much louder than others, use the Compressor, which turns down the loudest moments so the level is steadier (What Is Audio Compression? explains the idea). Its Vocal preset is a reasonable starting point, and you can adjust Threshold, Ratio, Attack, Release, Knee and Makeup gain while you preview. Skip it if the level is already steady. Compression brings up the quiet parts, noise included, so it also belongs after noise reduction.
5. Loudness last
Finish in the Volume Normalizer. Choose Loudness (LUFS) and pick a target (LUFS measures loudness the way ears hear it). For spoken word, -16 LUFS is a sensible choice (it's the target Apple Podcasts publishes), and the tool shows before and after readings.
Loudness goes last because normalizing and boosting are plain gain: they raise the noise floor along with the voice. Done first, the noise reducer gets louder noise and every later step shifts the level again. Done last, the number you set is the number you export.
The one-click shortcut
The Audio Enhancer runs a lighter version of this chain in a single pass. Pick Voice or Music, set Strength, and press Enhance audio. In Voice mode it filters rumble, shapes the EQ, compresses and sets loudness, which suits a recording that's only a bit rough.
The trade-off is control: it's a single pass, so you can't ease off one step while keeping the others. Its optional noise step is spectral, not the AI speech model, so it suits steady hiss and fan noise more than a busy room. For a properly bad recording, run the steps separately.
How to remove background noise from a video
Open the video file you already have in Video to Audio to pull out the soundtrack, then run it through the steps above, skipping the trim step so the audio keeps the video's length. In your video editor, mute the original audio and line the cleaned file up from the start of the clip, then check that the last few seconds still match the picture. A soundtrack longer than the Noise Reducer's 12-minute limit has to be split: clean each piece, then place each one back at the timestamp where it started.
What can't be fixed (and when to re-record)
Some damage is permanent, and pushing harder only adds new problems.
- Hard clipping. When a recording clips, the tops of the loudest waves are cut flat and that information is gone. A small cut at 6 kHz and 12 kHz can soften the edge, and you can trim the worst passages, but nothing restores the original.
- Heavy room echo. A distant voice in a reflective room is mixed with its own reflections. Voice mode reduces some of it, but the result is partial at best.
- Crosstalk and overlapping speakers. When two people talk over each other in one mixed file, there's no way to keep one voice and drop the other. A speech model treats both as speech.
- A signal buried in noise. When the voice is barely louder than the noise, the noise reducer can't reliably tell them apart, so it either leaves the noise or eats the voice.
- Over-processing. Too much noise reduction makes voices watery or robotic. A little remaining noise is easier to listen to than artifacts.
If a recording has two or more of these and you can record it again, re-recording is usually quicker than repairing it.
Prevent it next time
- Get the mic closer. Distance causes most echo, thin tone and noise problems at once. Mic Placement for Clean Voice Recordings covers distance and angle.
- Leave headroom. Keep the loudest moments well below the top of the meter, because clipping can't be undone.
- Pick a quiet, soft room. Turn off fans and AC, and record near curtains, carpet or furniture rather than bare walls.
- Do a short test. Record ten seconds and listen back on headphones before the real take.
Frequently asked questions
Can you fix a clipped recording? Not fully. Clipping happens when a signal is louder than the recorder can capture, so the peaks of the waveform are cut flat and the missing shape is lost. You can soften the harshness with a small EQ cut in the upper bands, trim the worst passages, or lower the level so it's less obvious, but no tool restores the original sound. Leaving headroom while recording is the only real fix.
What order should I apply noise reduction, EQ and normalization? Trim the file first, then run noise reduction, then EQ, then compression if the level is uneven, and set loudness last. Noise reduction goes early because EQ boosts and gain both raise the noise floor. EQ follows noise reduction because denoising changes the tone of the voice. Loudness goes last so no later step changes the level you set.
Can I remove background noise from a video? Yes. Extract the soundtrack from your video file with Vocal Cut's Video to Audio tool, clean it with the Noise Reducer and any other steps it needs, then go back to your video editor, mute the original audio, and line the cleaned track up from the start of the clip. If you didn't trim the audio during cleanup, it stays in sync with the picture.
Will noise reduction make my voice sound robotic? It can if you apply too much. Heavy noise reduction removes parts of the voice along with the noise, which leaves a watery, swirly or robotic sound. In Vocal Cut's Noise Reducer, the slider between Original and Denoised blends the two live, so you can pull it back until the voice sounds natural. A little remaining background noise is usually less distracting than the artifacts.
Is the free noise reducer private? Yes. Vocal Cut's Noise Reducer runs entirely in your browser on your own device, so your audio never leaves your device and is never uploaded to a server. Voice mode does download a small AI speech model to your browser the first time you use it, which is then cached. There's no account to create and no sign-up.