Saturalabs center-cancellation listening example Only these two original synthesized WAVs are dedicated under CC0 1.0: stereo-reference.wav center-cancelled.wav https://creativecommons.org/publicdomain/zero/1.0/ No other site media or software is covered by this notice. Source: three seconds, 48 kHz stereo PCM16. Both channels contain a 440 Hz sine component at amplitude 0.18. Left adds 660 Hz and right adds 880 Hz, each at amplitude 0.09. Five-millisecond linear ramps finish the source edges. The stored PCM16 source is read back and processed by the application's actual exportChannelWavs(..., "difference") function: (left - right) / 2. The result is a three-second mono PCM16 WAV with no extra fade or normalization. This demonstrates arithmetic on synthesized components. There is no recorded vocal, instrument performance, or AI output. It is not a song-quality benchmark. Quantization leaves a small residual; a centered instrument is not protected from subtraction just because a user wanted to remove a vocal. Reproduce from the repository with FFmpeg and FFprobe installed: node --import tsx scripts/generate-center-cancellation-example.ts Evidence: docs/experiments/center-cancellation-example.json. Independent FFmpeg decoding checks stored format, channels, duration, and all difference samples within one PCM16 quantization step. Exact-frequency amplitudes are measured over seconds 1-2. This does not test browser playback, speaker acoustics, or the audibility of particular components for a listener. These files are separate from the existing WAV test-pack ZIP.