Consistent structure. Honest limits.
The browser analyzer is a deterministic technical heuristic. It is not a generative hit predictor, listener panel, copyright determination, or calibrated commercial-success forecast.
Direct measurements
The local engine decodes the full file and estimates: sample peak; an approximate ungated K-weighted loudness level (not standards-compliant integrated LUFS) using an ITU-R BS.1770-style K-weighting filter cascade, applied ungated and with reference coefficients designed for 48kHz (used as a deliberate approximation at other sample rates); peak-to-RMS crest factor; a clipped-sample heuristic; L/R stereo correlation; a four-band spectral-balance split (bass, low-mid, high-mid, air) via cascaded low-pass filters; duration; sample rate; and channel configuration.
What the score means
The technical baseline index is a transparent, additive heuristic: measured values contribute continuously near broad risk boundaries, and each category's point change and measured evidence is shown in the report. Scores can still tie when tracks have similar measurements. It is not calibrated against a reference mastering corpus or validated listener outcomes, and it must not be treated as a quality grade. It must not be represented as the probability that a song will chart or become commercially successful.
Known limits
The loudness and ungated loudness figures are browser-computed approximations, not certified broadcast measurements — they are not produced by a calibrated BS.1770 meter, do not implement the standard's 400ms gated-block averaging, and are not validated against a reference test-vector suite. Treat them as directionally accurate, not lab-grade. This build still makes no claim about listener demand, cultural timing, playlist acceptance, radio response, or creative superiority.
Sixteen-section energy map & export
The energy map measures RMS dBFS in 16 equal time windows. It cannot identify verse, chorus, or emotion. Exported JSON contains the technical measurements, optional title and artist name, score rubric and limitations, but never audio samples.
Evidence firewall
Public chart rank must never silently alter direct waveform measurements. Creation provenance must never secretly raise or lower technical readiness. Human judgments should be labeled as judgments.
Regression requirement
This implementation has synthetic regression checks for score response and dynamic evidence text; that is not validation against real mastering references or chart outcomes. A validated golden audio library is still needed before claiming predictive validity.