YouTube-to-MIDI close-note test

Can YouTube to MIDI separate close notes?

Not reliably in this controlled test. When C4 and C-sharp 4 played together, the current converter kept C-sharp 4 but missed C4. It retained both notes in the wider major-third and octave pairs.

Test a local audio file

This is one synthetic two-note comparison, not a universal minimum interval or accuracy claim.

What we tested

We generated one 22,050 Hz, 16-bit mono WAV containing three two-note sections: C4 with C-sharp 4, C4 with E4, and C4 with C5. Each pair lasted 1.400 seconds and was separated by 0.500 seconds of silence. Every tone used the same sine waveform, peak level, timing, attack, and release. No public or named media was used.

IntervalGenerated notesExpected notes foundMIDI behavior
SemitoneC4 + C-sharp 41 of 2C4 missed; C-sharp 4 split into 4 events
Major thirdC4 + E42 of 2Both main notes plus 2 short tails
OctaveC4 + C52 of 2C5 continuous; C4 split into 4 events

The semitone collapsed to one pitch

The semitone section produced four C-sharp 4 events and no C4 event. Its longest C-sharp 4 segment lasted 1.092 seconds. The file therefore did not preserve both intended chord tones even though both source oscillators had equal peak amplitude.

Wider pairs kept both intended notes

The major-third section retained C4 and E4 with main durations of 1.347 and 1.358 seconds. The octave section retained C4 and C5; C5 lasted 1.358 seconds, while C4 was divided into four events. Across the complete file, the converter exported 13 events in a 161-byte MIDI file.

This comparison changes interval width and frequency while keeping the source construction fixed. It shows what happened to these three pairs, but it does not prove that wider intervals will always work or that every semitone will fail.

Why adjacent pitches are a harder review case

Audio contains overlapping frequency patterns rather than labeled MIDI keys. When two fundamentals are close, their combined waveform and harmonics can make separate note boundaries less clear to a transcription model. Instrument tone, tuning, phase, room sound, and other parts in a mix can all change the result.

How to review close harmony

  1. Check every intended chord tone. A plausible top note does not prove the lower adjacent note survived.
  2. Look for split events. Merge same-pitch fragments only after comparing them with the source.
  3. Try a cleaner stem. Isolating the relevant instrument can remove competing energy without changing the musical interval.
  4. Add missing notes manually. MIDI is an editable draft, so restore a known inner voice in the piano roll when needed.

Limits of this result

The test used steady equal-level sine tones and the converter's default controls. It did not test every register, interval, instrument, inversion, tuning, or chord density. It records one reproducible boundary from the current converter without assigning an accuracy percentage.

For ordinary triads, see the controlled chord test. For one note placed between equal-tempered pitches, see the detuned-note test.

After the note draft: MIDI Agent is a text-to-MIDI AI music plugin for a DAW that can help develop editable melody, chord, drum, and bassline parts from prompts.

Explore MIDI Agent