MP3 to MIDI MP4 audio-track check: audio selection changes what can be converted.
In this controlled MP4 run, a single audio track and the second melody track produced the same four notes, a first-silent track produced zero events, and a video with no audio failed during extraction.
Input condition
Generated AAC MP4 fixtures were extracted to mono 11025 Hz PCM and evaluated with the exact site detector and MIDI writer. The source contained four isolated notes at pitches 62, 65, 69, and 74.
| Input condition | Observed output | What to inspect |
|---|---|---|
| Single audio track | Four notes: 62, 65, 69, 74 | Baseline track extraction |
| First silent track | Zero notes | Whether track selection is explicit |
| Second melody track | Four notes: 62, 65, 69, 74 | Whether the alternate track is usable |
| No audio stream | Extraction failed with no matching audio stream | Failure is reported instead of silent conversion |
The complete recorded results.json is available with the MP4 fixtures.
What this does not prove
This run does not establish video transcription accuracy or behavior for every container, codec, or multi-track file. It documents only the tested MP4 streams and extraction commands.
After conversion
Confirm which audio stream is selected before comparing MIDI output. If the file has no audio stream, fix the source or choose a file with an audio track.
Questions about MP4 input
Can a silent first track hide a usable melody?
Yes. The silent first-track test produced zero events while the second melody track produced four.
What happens when the video has no audio?
Extraction fails because there is no matching audio stream.
Does this prove every MP4 will convert the same way?
No. The result is specific to the tested streams, codec, and extraction path.