What Is an Audio Limiter?

A compressor reduces gain above its threshold by a configurable ratio. 4:1 compression: for every 4 dB the input exceeds threshold, the output rises only 1 dB. A limiter uses very high ratios (10:1, 20:1, infinity:1) so the output cannot exceed the ceiling at all — every sample above the threshold is reduced to exactly the threshold value. A 'brick-wall' limiter guarantees the ceiling absolutely; a 'soft' limiter allows brief overshoots before clamping. Compressors shape the dynamic envelope of the audio; limiters cap peaks. Most mastering chains include both: a compressor 6-10 dB before the limiter for tonal density and dynamic control, then a brick-wall limiter at the very end for ceiling protection. See [fix-audio-clipping](/guide/fix-audio-clipping) for the related clipping topic.

Two different ceilings exist. Sample-peak limiting prevents any digital sample from exceeding the threshold — what you see on a peak meter looking at the file directly. True-peak limiting prevents reconstructed-waveform peaks from exceeding the threshold after D/A conversion or codec decoding. Inter-sample peaks can be 0.5-1 dB higher than the digital sample peaks because the smooth analogue waveform reconstructed by the DAC passes through values between sample points. A file mastered to 0 dBFS sample-peak can clip as -0.3 dBTP through a streaming codec. EBU R128, ATSC A/85, Apple Sound Check, and Spotify all require -1 dBTP true-peak ceilings. True-peak limiters oversample (4x typical) before measurement so they catch inter-sample peaks before they reach the consumer.

Three controls do most of the work. Threshold (sometimes called Input or Drive): the level above which the limiter engages. Lower thresholds engage more often, providing more loudness but more potential transient distortion. Ceiling (sometimes Output or Output Ceiling): the maximum allowed output level — set to -1 dBTP for streaming, -0.3 dBFS for CD masters, -1 dBFS as a safe universal ceiling. Release: how quickly the limiter recovers gain after a peak has passed; 50-200 ms typical. Fast release (10-20 ms) maximises loudness but introduces 'pumping' — audible level fluctuations on sustained content. Slow release (300-500 ms) sounds transparent but limits the peak loudness gain achievable. Modern limiters (FabFilter Pro-L 2, iZotope Ozone) include attack and lookahead controls; lookahead 5-10 ms gives the algorithm advance notice of incoming peaks for cleaner reduction.

Set the ceiling per delivery context. Spotify, Apple Music, YouTube Music, Tidal: -1 dBTP ceiling, content normalised to -14 LUFS. Apple Podcasts: -1 dBTP, normalised to -16 LUFS. EBU R128 broadcast (UK, EU): -1 dBTP, integrated -23 LUFS. ATSC A/85 broadcast (North America): -2 dBTP, integrated -24 LKFS. Cinema (Dolby Atmos consumer mix): typically -3 dBTP. Audiobook ACX: -3 dBFS sample-peak (older spec uses sample-peak rather than true-peak), RMS -23 to -18 dBFS. CD master: -0.1 to -0.3 dBFS sample-peak; the spec allows 0 dBFS but headroom protects against player downstream issues. When in doubt, -1 dBTP works for everything except cinema. See [audio-for-broadcasting](/guide/audio-for-broadcasting) for full broadcast specifications.

Limiting is destructive — every dB of peak reduction is a dB of dynamics permanently removed. Rules of thumb based on AES listening studies: 1-3 dB of gain reduction on peaks is essentially transparent on most material. 3-6 dB is audible on transient-rich content (drums, percussion, plucked strings) as a slight 'softening' of the attack but acceptable for streaming. 6-10 dB starts to dull transients noticeably and reduce dynamic range; mainstream pop masters operate here. More than 10 dB sustained is the loudness-war territory: pumping, distortion, and reduction in stereo imaging. The streaming-era trend is back toward less limiting — Spotify normalises -14 LUFS regardless of source, so over-limited masters are turned down on playback, eliminating the loudness benefit and keeping only the artefacts.

Industry standards: FabFilter Pro-L 2 (true-peak, 8x oversampling, 8 release modes, market favourite for transparent mastering). iZotope Ozone Maximizer (IRC IV intelligent algorithm, included in Ozone bundle). Waves L1/L2/L3 (the original digital brick-wall lineage; L3 is multiband). Sonnox Oxford Limiter (broadcast and TV mastering). Free options: TDR Limiter 6 GE — true-peak, transparent, the recommended free pick for mastering. LoudMax (LV-32) — minimal CPU, transparent. Ableton Live Limiter (built-in) — adequate for stems and check mixes. DAW stock limiters (Logic Adaptive Limiter, Pro Tools Maxim) work for non-mastering tasks. For broadcast loudness compliance: NuGen ISL 2, MasterCheck Pro — these include LUFS and true-peak metering plus codec preview to guarantee delivery spec.

Place the limiter as the very last insert on the master bus, after EQ, compression, and any saturation or stereo processing. Set the ceiling first to your target (-1 dBTP for streaming). Adjust input gain (or threshold) until the gain-reduction meter shows 2-6 dB on peaks for transparent results, or up to 6-10 dB for commercial-loudness pop. Choose a true-peak detection mode if available. Listen specifically to transient material — kick drums, snares, sibilants — for distortion or pumping. Compare loudness-matched against the unlimited mix; if the limited version sounds noticeably duller or smaller, back off the input gain. Verify with a LUFS meter that integrated loudness matches the target platform. Encode via [WAV to MP3](/convert/wav-to-mp3) at 192-320 kbps; the limiter's true-peak ceiling protects against codec inter-sample peaks. See related [audio-for-broadcasting](/guide/audio-for-broadcasting) and [audio-for-podcasters](/guide/audio-for-podcasters) for delivery specs.