What we tested
We generated three 22,050 Hz, 16-bit stereo PCM WAV files. Every file contained the same isolated C4, E4, G4, and C5 sine tones. Each tone lasted 0.800 seconds, with 0.300-second gaps and a 0.68 peak level in the active channel.
The in-phase file put the same signal in both channels. The left-only file left the right channel silent. The opposite-polarity file put the signal in the left channel and its sign-inverted copy in the right. We ran all three through the current local converter at its default controls and inspected the mono buffer and exported MIDI.
| Stereo condition | Mono peak | MIDI events | Detected pitches | MIDI size |
|---|---|---|---|---|
| Same signal in both channels | 0.68002 | 7 | C4, E4, G4, C5 | 113 bytes |
| Signal in left channel only | 0.34001 | 7 | C4, E4, G4, C5 | 113 bytes |
| Right channel polarity inverted | 0.0000105 | 4 | C5, E5, G5, C6 | 89 bytes |
What changed after the stereo-to-mono step
The decoded left and right channels of the opposite-polarity file had a correlation of approximately -1. After the browser rendered them to the mono buffer used for transcription, the measured peak fell from about 0.68 in either source channel to about 0.0000105.
The result was not a clean “no notes” message. The remaining tiny signal produced four stable events, but every pitch was one octave above the source: C5 instead of C4, E5 instead of E4, G5 instead of G4, and C6 instead of C5. All three runs returned HTTP 200 and produced no console errors.
Why this differs from ordinary stereo placement
The earlier stereo test put different notes in the left and right channels. Both sides contributed to a single mono note draft, and their original pan positions were discarded. This test uses the same signal on both sides but reverses one channel's polarity, so the two waveforms oppose each other during downmixing.
That distinction matters: hard-left or hard-right placement alone did not remove notes in the earlier test. The failure here depended on a deliberately matched, sign-inverted pair.
How to check a stereo source before conversion
- Listen in mono. If the instrument becomes much quieter or changes character, channel interaction may affect transcription.
- Try one channel by itself. In this test the left-only version kept the intended pitches even though its mono peak was half the in-phase peak.
- Fix the source, not the MIDI guess. Correct an accidental polarity inversion or export a clean single channel before converting.
- Verify octaves against the recording. A tidy-looking MIDI result can still be musically wrong.
What this test does not prove
It does not establish a phase threshold, predict partially shifted microphones, or measure real recordings, compressed video audio, instruments, mixes, or other converters. It records the current tool's exact output for three controlled files and shows why mono compatibility can matter before YouTube-to-MIDI transcription.