Limiting
The final dynamics stage — how limiters work, when to use them, and how loud is too loud.
A limiter is a compressor. That's it. There's no magic, no separate category of device — just a compressor with an extreme ratio. Where a typical compressor might use 4:1 or 8:1, a limiter uses ∞:1. Nothing gets past the ceiling.
Understanding that single fact clears up most of the confusion around limiting.
Specs
Limiting vs Compression
Both are dynamics processing. Both reduce the level of signals that exceed a threshold. The difference is degree.
A compressor at 4:1 still lets transients through — they're reduced, but not stopped. A signal 8 dB over threshold comes out 2 dB over. That's shaping. You're sculpting the dynamic range whilst keeping the feel of the performance.
A limiter at ∞:1 stops everything at the ceiling. A signal 8 dB over threshold comes out at the threshold. Full stop. There's no proportional reduction — there's a wall.
In practice, most mastering limiters operate at ratios somewhere between 10:1 and ∞:1, with extremely fast attack times. The result is the same: nothing meaningful gets past the ceiling you set.
Where compression shapes dynamics, limiting prevents peaks from exceeding a level. Compression is an artistic choice. Limiting is a technical boundary.
How Modern Limiters Work
A basic limiter — fast compressor, infinite ratio — works, but it's crude. Modern mastering limiters are significantly more sophisticated.
Lookahead. The limiter reads a few milliseconds ahead of the audio, spotting transients before they arrive. This lets it engage gain reduction smoothly rather than reacting after the peak has already hit. The result is cleaner, more transparent limiting. Most mastering limiters use 1–5 ms of lookahead by default. The tradeoff is a tiny amount of added latency — irrelevant during mastering, but worth knowing about.
Oversampling. Digital audio is a series of sample points. Between those points, the actual waveform can peak higher than any individual sample suggests. These are inter-sample peaks (ISPs), and they're invisible to a standard peak meter. Oversampling runs the limiter at 2×, 4×, or even 8× the session sample rate, catching those hidden peaks before they clip the output. This matters enormously for streaming — more on that below.
Multi-band limiting. Rather than applying a single threshold across the entire frequency spectrum, multi-band limiters split the signal into frequency ranges and limit each independently. A loud kick drum won't cause the high frequencies to duck. This preserves clarity and prevents the pumping that single-band limiters can introduce on complex material.
Mastering Use
Limiting is typically the final processor in the mastering chain — after EQ, after compression, after everything else. Its job is straightforward: set the maximum output level and push the overall loudness up to the target.
Two controls matter most:
Ceiling. The absolute maximum output level. For streaming platforms, set this to −1.0 dBTP (true peak). For CD, −0.1 dBTP is standard. The ceiling exists to prevent clipping downstream — in codecs, in playback systems, in the listener's ears.
Threshold (or Input Gain). This determines how much gain reduction the limiter applies. Pull the threshold down (or push the input gain up) and the limiter works harder, producing a louder output. The tradeoff is always the same: more loudness equals less dynamic range.
That tradeoff is real and unavoidable. There is no free loudness. Every dB of gain reduction is a dB of dynamic range you've removed from the music. The question isn't whether to limit — it's how much limiting the material can tolerate before it starts sounding worse.
Platform Loudness Targets
Every major streaming platform normalises loudness. If your master is louder than the platform target, it gets turned down. If it's quieter, some platforms turn it up. This means mastering significantly louder than the target gains you nothing — you've just crushed your dynamics for no benefit.
The key targets:
- Spotify: −14 LUFS (integrated)
- Apple Music: −16 LUFS (integrated)
- YouTube: −14 LUFS (integrated)
- Amazon Music: −14 LUFS (integrated)
- Tidal: −14 LUFS (integrated)
These are integrated loudness measurements — the average over the entire track, not the peak. A master at −14 LUFS integrated with a −1.0 dBTP ceiling is the sensible default for most streaming-focused releases.
True Peak vs Sample Peak
This distinction matters more than most people realise.
Sample peak is the highest sample value in your audio file. It's what a standard peak meter reads.
True peak is the highest level the reconstructed analogue waveform reaches between samples. It can be — and frequently is — higher than the sample peak. A file that reads −0.3 dBFS on a sample peak meter might actually hit +0.8 dBTP when reconstructed.
Why does this matter? Because streaming platforms encode your audio with lossy codecs (AAC, Ogg Vorbis, Opus). The encoding process reconstructs the waveform, and those inter-sample peaks become real peaks. If your true peak level is above 0 dBTP, the codec output clips. The listener hears distortion that wasn't in your master.
This is why mastering for streaming means setting a true peak ceiling, not a sample peak ceiling. −1.0 dBTP is the standard recommendation. Some engineers use −0.5 dBTP. Going right to 0 dBTP is asking for trouble.
Use a limiter with true peak detection enabled. Every serious mastering limiter offers this. There is no good reason to leave it off.
Common Mistakes
Slamming the limiter too hard. If you're seeing 6+ dB of gain reduction on every beat, you're past limiting and into distortion. The material pumps, transients smear, and the low end turns to mush. Back off. If the track isn't loud enough with moderate limiting, the mix needs work — not more limiting.
Ignoring true peak. Mastering to a sample peak ceiling of 0 dBFS and calling it done is how you get codec clipping on every streaming platform. Always use true peak metering. Always.
Mastering louder than the platform target. Your track at −8 LUFS will be turned down to −14 LUFS on Spotify. It'll sit at the same perceived loudness as a track mastered to −14 LUFS — except yours has half the dynamic range. You've gained nothing and lost everything. Master to the target, not past it.
Using limiting as a substitute for good mixing. A limiter cannot fix a mix with poor balance. If the vocal is too quiet, the limiter will make everything louder — including the things that are already too loud. Fix the mix first. Limit last.
Setting the release too fast. An overly fast release causes distortion on low-frequency material. Too slow, and the limiter can't recover between transients, causing audible pumping. Most modern limiters have intelligent release algorithms that handle this automatically. Use them.
Recommendations
The modern mastering limiter. Multiple algorithms, true peak, loudness metering.
View →Not a limiter, but essential for seeing your LUFS. Use it alongside your limiter.
View →The classic. Simple, effective, on countless masters.
View →