sample*output
four pages · rendered from flirting virtual orchestra
[ back to fradan ]
[ back to the void ]

These are the four pages exactly as fradan renders them. The track is flirting virtual orchestra, one of the void machines tracks, decoded from its lossless master.

Because the source is a lossless, page one comes back close to clean. The method looks for a codec's fingerprint, and there is no codec in the chain. This is a track that is using physical modelling, which is obvious if you pay close attention to pages 2 and 3, page 4 purpose is to assert the mastering of a track

PAGE 1:: CODEC ARTIFACTS
Spectrum over time with the per-frame lowpass cutoff traced over it, the artifact distribution showing only what was removed, localised per-second evidence, cutoff distribution, holes by band and the mean shelf profile.
PAGE 2:: SYNTHESIS VS CAPTURE
Stereo coherence and width over frequency across five time slices, noise floor against level, modulation spectrum by band, and top-octave activity against the whole band.
PAGE 3:: TIMING AND TRANSIENTS
Local tempo over time, every onset's distance from a fitted sixteenth-note grid, spectral seams, the distribution of 10 to 90 percent attack rise times, and the average attack shape aligned at its peak.
PAGE 4:: LEVEL, DYNAMICS AND BALANCE
Momentary and short-term LUFS against the gated integrated value, the short-term loudness histogram with its p10 to p95 range, peak to short-term crest, third-octave long-term average spectrum, whitened resonances, and a goniometer with inter-channel correlation.
what to look for
[ lowpass shelf ]
an encoder's brickwall, traced per frame
[ spectral holes ]
bands quantised to zero inside the passband
[ stopband floor ]
a codec leaves near-silence, nature leaves noise
[ stereo coherence ]
two mics in a room decorrelate with frequency
[ noise floor slope ]
room tone is flat, gating bends it
[ grid lock ]
whether one step fits the whole file
[ attack rise time ]
struck sounds arrive in a couple of ms
[ loudness range ]
p10 to p95 of the short-term values
[ analyse your own file ]