A full-reference check of what a lossy codec did to your audio: upload the original master and the encoded file you want to inspect (an MP3, AAC or Opus download, or your own encode), and the detector aligns the two in time and level, then reads the residual (original minus decoded encode) the way codec engineers audition it. Four artifact families are measured with the mechanisms described in the AES perceptual-audio-coding literature: PRE-ECHO — block-based coders quantize a whole analysis block, so the quantization noise of a transient spreads backwards and “announces” the attack; the tool finds sharp attacks that rise out of a quiet lead-in and reports how much noise the encode added in the 30 ms before each onset. TRANSIENT SMEARING — the attack sharpness of each transient is compared between original and encode, so a slowed, rounded attack shows up as attack loss in dB. BANDWIDTH LOSS — long-term average spectra of both sides give each an effective bandwidth (50 dB down from spectral peak); lossy encoders low-pass their input, and the loss in Hz plus the third-octave band deltas show where the top end went. STEREO DAMAGE — side-channel (L minus R) energy and the inter-channel correlation are compared, exposing joint-stereo collapse. The report grades every family (none / minor / notable / severe) and lists the worst events with timestamps so you can jump straight to the damaged spots in your DAW. These are energy-based indicators on the decoded signal, not a PEAQ/ODG perceptual score: they tell you where to listen, and encoders differ, so verify by ear before rejecting a delivery.