Measure your track, then master it — right here
Drop in a file or a link and the lab decodes it, measures it for real, and renders a mastering chain you control stage by stage. Loudness targeting, a true-peak limiter and WAV/MP3 export included. Nothing leaves this tab.
1 · Load a track
No track loaded yetDecoding uses your browser's own decoder. If a format cannot be decoded here, the lab says so instead of guessing.
2 · Analysis
Waiting for audioBand balance
Waveform with the detected energy sections. Click anywhere on it, or a section chip, to play from there.
Loudness, true peak, dynamic range, noise floor and tempo are estimates computed in this tab from the decoded PCM — close enough to work with, not a certified meter.
4 · A/B and spectrum
Nothing playing| Metric | Before | After |
|---|---|---|
| Loudness (est.) | — | — |
| True peak (est.) | — | — |
| Crest factor | — | — |
| Phase correlation | — | — |
| Stereo width | — | — |
| Normalization gain | — | — |
| Limiter reduction | — | — |
Average spectrum, log frequency: grey is the original, purple is the mastered render.
3 · Mastering chain
Not rendered yetClears sub-20–40 Hz rumble that eats headroom without being heard.
Narrow notches on the mains frequency and its second harmonic.
One control, two complementary shelves: negative warms the balance, positive opens the top.
A narrow cut around 3–5 kHz that only engages while that band is actually loud.
Gentle glue at a soft knee — a couple of dB of movement is usually plenty.
Mid/side scaling. Watch the correlation row after rendering: below 0 the mix will lose energy in mono.
The render is measured, then gained so the integrated estimate lands on your target before limiting.
Look-ahead limiter driven by a 4× oversampled peak detector, so inter-sample peaks stay under the ceiling too.
Applied only when the export reduces bit depth (16-bit WAV or MP3). One LSB of TPDF noise, deterministic per session.
Metadata (standard tags)
No file read yet5 · Export
Nothing exported yetThe result line reports the exported file size and the measured loudness and true peak before and after the chain.
What this lab is, and is not
Are these numbers certified measurements?
No. The loudness figure is a K-weighted, gated estimate in the spirit of BS.1770, and the true peak comes from 4× oversampling. They are accurate enough to master against, but a compliance meter they are not — every such value is labelled as an estimate.
Does anything get uploaded?
No. Decoding, measurement, rendering and encoding all happen in this tab. The only network call is the optional proxy fetch when you paste a link, and that is just to get the bytes past CORS.
Can it remove watermarks or provenance data?
No, and that is deliberate. This is an audio engineering tool: it measures and processes sound. It does not detect, strip or obscure watermarks or C2PA manifests, and re-rendering audio here is not a way to make a track unidentifiable.
Why does the mastered version sound louder in A/B?
Because it is, unless you keep the loudness-matched box ticked. Matched compare gains the original up to the mastered loudness so you judge the processing rather than the volume.