Audio Fake Detector PRO

Audio Fake Detector PRO analyses whether an audio signal genuinely contains the frequency content implied by its container bitrate.

Files are processed using FFmpeg (decoding to standardized PCM) and ffprobe (metadata extraction such as codec, bitrate, and stream parameters) as a preprocessing step.

It is designed to detect fake high-bitrate lossy files, such as MP3 audio upsampled or re-encoded to appear as 320 kbps, as well as lossy material (MP3/AAC) placed inside lossless containers (FLAC, WAV) to simulate higher quality.

The signal is split into multiple non-final analysis segments plus one terminal segment. Each segment is analysed independently via STFT-based spectral inspection. A majority-based voting system aggregates segment-level results into a single file-level classification. Lossless files are additionally validated through statistical PCM analysis using auCDtect.