ORIGINLAB / DETECTOR
AI music forensics
Inspect a track from several angles.
The quick engine combines a vocoder fakeprint, file provenance clues, and transparent spectral, timing, dynamics and stereo diagnostics. Add SONICS later for a heavyweight second opinion.
LOCAL SESSIONNo third-party audio API
STEP 01
Select source audio
Drop a track here, or choose a fileWAV, FLAC, MP3, M4A, AAC, OGG or Opus · up to 300 MBAudio stays on your private engine
Usually 15–90 seconds Fast ONNX profile
ANALYSIS STACK
Signals in this build
01
Vocoder fakeprint
MISSINGTiny ONNX classifier over periodic 1–8 kHz reconstruction residue.
02
Metadata & provenance
READYContainer tags, generator strings and manifest-presence inspection.
03
Transparent diagnostics
READYSpectral, onset, dynamics, stereo and window-level readouts.
04
SONICS ensemble
OPTIONALOptional SpecTTTra models for a style-based second opinion.
“Not detected” is not the same as proven human. Keep the signal-level evidence with every verdict.