Why measure integrated LUFS with ITU-R K-weighting instead of peaks?
Simple peak boosting reacts only to one stray transient spike, leaving quiet dialogue underpowered. Measuring integrated LUFS applies perceptual frequency filtering and 400 ms gated blocks under ITU-R BS.1770-4, letting you level quiet recordings to exact targets locally without harsh distortion.
Pre-warped K-weighting stages across 400 ms blocks
Loading a recording or clicking Load sample quiet audio (sample-quiet-podcast.wav at -28 LUFS) decodes tracks at 8,000 Hz to 192,000 Hz into planar Float32Array channels under the W3C Web Audio API 1.1 specification. Channel weights assign 1.0 to front channels, 0.0 to LFE, and 1.41 to surround channels.
designKWeightingStages applies a pre-warped +4 dB high-shelf filter near 1,682 Hz followed by a 38 Hz RLB high-pass filter before summing energy across 400 ms blocks stepped every 100 ms. Before leveling dialogue, you can pace spoken narration without shifting pitch inside the same audio workspace.
Two-stage -70 LKFS and -10 LU relative gating
Long silences between spoken phrases would pull down a naive un-gated average. First, measureIntegratedLufs applies an absolute -70 LKFS gate, dropping silent intros and dead air from the block mean.
Second, a -10 LU relative gate removes blocks quieter than ten loudness units below the first-stage mean before reporting Input integrated LUFS. Alongside integrated loudness, the input cards display Short-term max LUFS across sliding 3.0-second windows, Input true peak, and Input sample peak.
Preventing clipping with 4x Catmull-Rom true-peak limiting
Lifting a -28 LUFS interview to -14 LUFS applies +14 dB of linear gain, which pushes brief consonants or snare hits above digital full scale. Oversampled true-peak detection and proactive lookahead limiting eliminate digital clipping cleanly.
Catching inter-sample overshoots via 4x Catmull-Rom
Discrete samples read from an MDN Web Audio AudioBuffer interface can remain below 0 dBFS while the reconstructed analog curve between points spikes higher. Evaluating four-tap Catmull-Rom FIR interpolation at 0.25, 0.50, and 0.75 fractional sample offsets catches those hidden inter-sample peaks accurately.
Both dBTP true peak and dBFS sample peak display immediately upon loading so hot masters are obvious before processing. When combining separately mastered podcast segments, you can combine leveled chapters into one episode with unified gain.
Running an O(N) 5 ms lookahead ceiling limiter
After scaling channels by targetLufs - inputLufs (clamped from -36 dB to +36 dB, or 0 dB when input sits below -69.5 LUFS), applyLookaheadTruePeakLimiter inspects every sample’s 4x peak against True-peak ceiling. A monotonic deque computes the minimum required gain across a 5 ms lookahead horizon in O(N) time.
Rather than flattening transients into harsh square waves, 0.8 ms attack and 65 ms release smoothing envelopes dip gain right before loud impacts. Post-limiter re-measurement then updates Output LUFS, Applied gain (dB), and Output true peak (dBTP) accurately.
Choosing streaming, podcast, and broadcast LUFS targets
Picking the right Target loudness preset aligns your file with platform playback engines so streaming apps neither turn your track down nor leave it buried. Standardized 16-bit PCM output headers align with the Library of Congress Broadcast WAVE format specification for predictable broadcast ingestion.
| Target loudness preset | Integrated target | Default ceiling | Distribution standard | Loudness behavior |
|---|---|---|---|---|
| Streaming (-14 LUFS) | -14 LUFS |
-1.0 dBTP |
Spotify, YouTube, Tidal | Raises quiet demos to web streaming levels |
| Podcast / Apple (-16 LUFS) | -16 LUFS |
-1.0 dBTP |
Apple Podcasts, stereo mobile | Balances vocal clarity with healthy dynamics |
| EBU R128 (-23 LUFS) | -23 LUFS |
-1.0 dBTP |
European broadcast delivery | Leaves wide headroom for dynamic mixes |
| ATSC A/85 (-24 LUFS) | -24 LUFS |
-2.0 dBTP |
US broadcast television | Matches dialnorm compliance for TV dialogue |
| Manual gain (dB) | -24 to +24 dB |
-1.0 or -2.0 dBTP |
Custom stem boost or cut | Applies fixed linear gain plus ceiling limiter |
Multi-segment recordings whose speakers vary drastically between halves gain extra consistency when split prior to normalization. Use silence-based audio chapter splitting to divide a long show first so each individual track hits your chosen -16 LUFS loudness cleanly.
Tip: Select -2.0 dBTP under True-peak ceiling when mastering loud
-14 LUFStracks destined for128 kbpsMP3 or AAC encoders, because lossy codec quantization often adds0.5 dBto1.2 dBof extra peak overshoot.
Exporting normalized WAV, M4A, and WebM files offline
Once Normalize loudness finishes, a 160-bar Normalized audio waveform updates above an inline HTML5 player alongside the Normalized WAV byte count. Every normalized file saves directly from client RAM with the preset suffix appended (-lufs-14, -lufs-16, or -plus12db).
Saving 16-bit WAV, AAC, or Opus audio exports
Clicking Download normalized WAV encodes lossless 16-bit integer linear PCM RIFF WAV bytes directly from worker buffers. On browsers supporting the W3C WebCodecs specification AudioEncoder, Download .m4a (128 kbps AAC) and Download .webm (128 kbps Opus at 48,000 Hz) buttons appear beside WAV.
Keep uncompressed WAV whenever feeding normalized stems back into desktop DAWs, video timelines, or mastering archives without generational loss. Pick M4A or WebM when publishing compact final episodes to podcast feeds, mobile players, or browser embeds. Encoded MP4 and WebM containers preserve channel layout and 48,000 Hz sample metadata cleanly inside local memory.
Keeping unreleased masters and memos in browser RAM
Private board meetings, unreleased album tracks, and personal voice recordings never cross the network during K-weighting measurement or peak limiting. Both audioMeasureLufs and audioNormalize run inside everyday.worker.ts so heavy buffer math stays isolated from UI sliders.
Creators boosting short chimes or notification hooks can immediately trim them into phone-ready loops with our Ringtone Maker right after leveling. Temporary Blob URLs backing the output player revoke automatically when re-rendering or closing your tab.