Every producer has a folder like this: a kick with a click that cuts through everything, a vocal chop from an old record, a snare lifted from a free pack. Half of them sound great. The other half have hum, hiss, mud, and a resonance that fights the mix.
Cleaning them up is not glamorous. It is also the difference between a sample that sits and one you fight for an hour. I keep a small chain I run things through before anything goes in a project, and it costs nothing.
Know what you’re actually fixing first
Before touching a plugin, listen to the sample soloed, loud, more than twice. Bad samples fail in predictable ways:
- Noise floor: hiss, hum, computer fan, room tone. Constant, sits underneath everything.
- Clutter: a peaky kick, harsh high mids, boxy low mids.
- Sibilance or transients: a vocal “S” that pokes out, a cymbal that bites too hard.
- Tone imbalance: the sample is just coloured wrong for your track.
Each one wants a different tool. Reaching for EQ on everything is why most cleanup sounds worse than the original.
Gating first, not EQ
If the problem is a hiss or hum that only shows up in the gaps, a gate does more than any EQ move. Insert it, set the threshold just above the noise, and time the release so it doesn’t chop the tail. For free options, the Insta-Gate Noisegate from the Parish Audio store is a good place to start, and it’s free.
One caveat: gates reveal their work on sustained material. On pads, reverb tails, or anything with a long decay, a gate will pump and sound worse than the hiss. If that happens, don’t fight it, move to a gentle broadband tool instead.
Do the boring EQ move before the fun one
I start with what working mastering engineers call a cleaning EQ: not shaping, just turning down what’s too loud. High mids that bite, low mids that box, a resonant peak that rings on every hit.
The key is restraint. Broad Qs, small moves. If you’re notching with a razor-thin Q across six frequencies, you’re not cleaning, you’re redesigning the sample and it will sound hollow. I use a clean, transparent EQ for this pass specifically so I’m hearing the sample, not the plugin’s character.
Two to four moves max. If a sample needs more, it’s the wrong sample. Bin it and dig for another one.
When a de-esser beats an EQ
EQ is static. It turns a frequency down permanently. A de-esser turns it down only when it gets loud. On a vocal with poking “Ss”, or cymbals that spike on certain hits, that difference matters. You can do it with EQ, but a de-esser does it better when the problem is occasional.
The free S1 De-Esser from Parish Audio is a solid fixture for this, available as a VST3 for macOS and Windows. Two features make it genuinely useful for sample cleanup rather than just vocal work. The Range control caps maximum reduction, which stops a sample from going dull or lispy when a few hard consonants hit. The Monitor function lets you hear what’s being removed, so you can verify it’s targeting actual sibilance and not shaving off important detail.
That monitor mode is the single best way to learn what a de-esser is doing. Use it once and you’ll stop abusing the plugin.
One thing to remember from the mastering side: de-essing the whole track should stay subtle. You’re treating one layer of a mix, not a solo piece.
Sat it, don’t scrub it
Sometimes a noisy or thin sample is better off saturated than cleaned. Tube-style saturation squashes dynamics and adds harmonics that mask a lot of trash in the upper range. Free packs like Samplephonics’ Tube Drums do this for you, and they include patches for Kontakt, Ableton Sampler, Reason NN-XT and SFZ.
Practical rule: if a sample has ugly top end but good body, saturation hides the ugliness and gives you character. If the sample has a low hum or a DC offset, saturation makes it worse.
Compress last, and lightly
Only after cleanup. A light glue-style compressor evens out dynamics and pulls the sample together. Heavy compression on a badly cleaned sample just amplifies the damage. If it sounds worse after compressing, undo it, back off the ratio, and re-check the EQ.
Legal side, in one paragraph
If you grabbed the sample from Splice, Loopmasters, Cymatics, LANDR Samples, or Output Arcade, you’re on pre-cleared ground, but read the license. Some libraries require a license once your song hits a streaming threshold. And even with royalty-free loops, don’t use them exactly as-is. Chop, flip, layer. Partly for uniqueness, partly so nobody else’s track sounds like yours.
If you pulled it from a record, YouTube, or TikTok, you’re in clearance territory: masters owner, composition owner, permission, possibly fees. Tracklib handles pre-priced licensing. Rights holders can be reached via publishers or the Harry Fox Agency. “Too small to notice” is not a real category anymore. Indie tracks go viral and get demonetized every year.
A short chain I actually use
- Gate (if noise is intermittent)
- Cleaning EQ: two to four broad moves
- De-esser (only if sibilance or harsh spikes)
- Saturation (only if the sample is thin or the top end is ugly)
- Light glue compression
- Bypass everything and A/B once more
That last step is the one most people skip. Bypass often. If the “cleaned” version is not obviously better, keep the original and move on.
Also consider a Rename as you go. Files like kick_splice_01.wav and kick_splice_01_clean.wav sit next to each other forever. Pick a suffix you’ll remember, or you’ll redo the same work in six months.
What’s the worst noise problem you’ve had to fix in a sample? The one that made you almost give up on it.