What a mastering limiter does, and what we got wrong in ours
updated · November 2, 2026 · by Alex Pritsert
A mastering limiter is the last stage in the chain. It stops peaks from rising above a set ceiling, which lets you raise the overall loudness without digital clipping. A good one turns the gain down smoothly just ahead of each peak and recovers without being heard. A careless one adds distortion you can measure.
Short answer
A limiter is a compressor with a hard ceiling. You set the maximum level, often -1 dBTP, and it lowers the gain whenever a peak would pass it. Turn the input up and the track gets louder while the peaks stay put, at the cost of some transient punch.
The quality depends on how the gain moves. Lookahead lets the limiter start reducing before a peak arrives, and release sets how quickly it recovers. True-peak detection catches levels between samples that a basic meter misses. We learned how much this matters by measuring our own: the limiter in Tunr 0.4.5 changed gain in instant steps, which added about 80 dB of energy above 4 kHz on a test signal that had almost none. We replaced it in 0.5.0.
What a limiter does to your track
This article is for people who have pulled a limiter's input up, heard the song get louder and then wondered what they just paid for it. It sits at the end of the mastering chain, looks at the signal and calculates how much it must turn down so nothing crosses the ceiling. Most of the song passes untouched. The snare hits and kick transients get turned down for a few milliseconds each.
The amount of gain reduction is the number to watch. A couple of decibels on the loudest hits usually goes unnoticed. As you push further, the limiter is working almost constantly, transients flatten and the song starts to feel smaller even as it measures louder.
Attack, release and lookahead in plain words
Attack is how fast the gain goes down, and release is how fast it comes back. Too fast a release adds distortion on bass notes, because the gain starts following the waveform itself. Too slow and the track stays turned down after a hit, so the next few notes sound quieter than they should.
Lookahead solves a timing problem. A limiter must lower the gain before a peak arrives, so it delays the audio by a few milliseconds and uses that time to ramp smoothly. A dual release has two speeds: a quick one for isolated transients and a slower one for sustained loud passages.
What is true peak, and why is the ceiling -1 dBTP?
Digital audio stores samples, but your speakers play the continuous wave reconstructed from them. That wave can rise higher between two samples than either sample shows, so a meter reading only sample values can under-report. A true-peak meter estimates those in-between levels by oversampling.
The ceiling exists because of what happens next. EBU R 128 sets a maximum of -1 dBTP for broadcast programme, and Spotify asks for true peak below -1 dB, or below -2 dB when a master is louder than -14 LUFS, to avoid extra distortion during encoding. In Tunr's codec preview, MP3 and AAC commonly land half a decibel to a decibel above the ceiling set before encoding.
Tunr's default ceiling is -1 dBTP, and you can lower it. If you're delivering a loud master to Spotify, use its -2 dB advice and check the result in the codec preview.
What we found when we measured our own limiter
We audited Tunr 0.4.5 in October 2026 and found four defects in the signal path. The first was the limiter, which changed gain in instant steps, with no ramp. To measure the damage, we fed it a test signal containing only 200 Hz and 1 kHz, in 30 ms bursts reaching 0 dBFS, driven 6 dB into the limiter. Any energy above 4 kHz had to be something the limiter added.
Between 4 and 20 kHz, energy rose from -142.9 dB at the input to -62.6 dB at the output, a gain of 80.3 dB. That energy sat about 60 dB below the programme and clustered at the edges of the bursts. You might hear it as faint clicks on sparse material, while a dense mix would hide it. We had shipped that limiter and didn't catch it until the audit.
The glue compressor had a related fault. With a 30 ms attack requested, it reached its full 4 dB of gain reduction within 0.5 ms, so the setting did nothing and every transient was clamped. Meanwhile the loudness stage was a gain boost into a soft clipper aimed at -7 LUFS by default, and saturation ran without oversampling, which caused aliasing on bright material.
| Defect in 0.4.5 | What we measured or found | 0.5.0 |
|---|---|---|
| Limiter gain moved in steps | 80.3 dB added above 4 kHz on the test signal | Lookahead, dual release, 4x oversampled detection |
| Compressor ignored attack | Full reduction by 0.5 ms with 30 ms requested | Attack and release behave as labelled |
| Loudness stage | Gain into a soft clipper, -7 LUFS default | Balanced preset targets -11 LUFS by default |
| Saturation | No oversampling, aliasing on bright material | Oversampled saturation |
How to tell whether a limiter is hurting your master
You don't need a lab. Bounce the same section with and without the limiter, match their loudness, and compare them. Matching matters most, because the louder version will tend to win otherwise. Listen to the snare and the vocal consonants first. If the limited version loses its front edge, back off.
A null test goes further: flip the polarity of one version and play them together, so only the difference remains. Tunr's A/B Mixer has a null test, a loudness match and a blind test for this. Each defect above now has its own test in our engine too. On the burst signal, the new limiter must hold high-frequency energy at or below -75 dB, and it must beat the old one by at least 12 dB.
Very loud targets have their own warning. Tunr's Dense and Club presets can stop short of the target when reaching it would hold more than 30% of the samples in gain reduction. The report says by how much, and adding the compressor or multiband stage before the limiter usually closes the gap. For choosing a number in the first place, see how loud a master should be.
Frequently asked
What is the difference between a limiter and a compressor?
A compressor reduces level gradually above a threshold by a ratio, shaping the dynamics of the track. A limiter works like a compressor with a very high ratio and a fixed ceiling, so peaks can't pass it. Mastering chains often use both, with the limiter placed last to catch what remains.
How much limiting is too much?
There is no fixed number, because genre and source vary. A couple of decibels on the loudest hits usually goes unnoticed. Compare the limited and unlimited versions at matched loudness. When the snare loses its front edge or vocals start to feel strained, you have gone too far, whatever the meter says.
Should my true peak ceiling be -1 or -2 dBTP?
Spotify asks for true peak below -1 dB, and below -2 dB when the master is louder than -14 LUFS, to avoid extra distortion during encoding. So -1 dBTP suits masters at or below -14 LUFS. For louder masters, -2 dBTP gives lossy encoders more room and is the safer choice there.