I watched this Youtube video about converting audio to MIDI, and the comments mostly point out that it's a scam or hoax.

Is it really difficult for software to convert audio to MIDI, and if so why?

Also, do pro musicians find this hard too?

  • I've read about some software tools that have made large advances in converting audio to MIDI, but several converter tools transcribe unpitched percussion as pitched notes, with disastrous results.
    – Dekkadeci
    Commented Dec 11, 2018 at 15:35

2 Answers 2


Yes, it's hard. And it gets harder as the music gets more complex. I think there would be general agreement that, at the present state of the art, a skilled and experienced musician is way ahead of any computer program. But computers have a habit of catching up....


I had a quick listen to part of the YouTube video - they guy is confusing audio & MIDI a bit, really. He's done the multi-pitch analysis, & is then dragging notes up & down... fine, but that's not MIDI, that's Melodyne live-shifting one single note from the chord as audio, in realtime [which is one hell of a feat anyway]

To the uninitiated - Melodyne is "Auto Tune" on steroids.

It is neither scam nor hoax. I own the software & have done for a decade [or more] The manufacturer, Celemony, have a scant entry on Wikipedia but you can investigate Melodyne on their own site.

Melodyne is about the only commercially-available software than can analyse chords & separate the pitches. With audio, you can then shift pitches within the file without changing the entire song.

Whether it can convert this, without error, into MIDI would be a whole different level of task, & one I have never actually needed to test. I use it entirely as an audio medium.

What it can do is such as this...

REM - Losing My Religion in a major key

Yes, it sounds truly weird :\ but if you didn't know the original, I doubt you would be able to tell what had been done to it.

Pitch tracking itself has come a long way over the years.
Back in the late 90's I worked for Yamaha R&D & we were truly proud at that time to be able to play midi drums & piano, whilst live-tracking a real bass guitar, audio to midi, & sync that all to a streamed vocal recording, with a video of the entire performance, as a kind of 'simulcast'.
It's too old to have ever made it to YouTube, so I can't demonstrate how it did actually work - even in a rather primitive way.

Do pro musicians find this hard too?
Some do, some don't. I know many people who can transcribe accurately with little repetition, though mostly they listen to one instrument at a time.
I do know one guy who can transcribe an orchestra, vertically, down the page, in 2-bar segments, pausing the audio only long enough for him to physically write all the notes.
That is an extremely rate talent.

I realised I have one of his transcriptions already on Soundcloud - this is a MIDI file, a "one box demo" i.e. all played at once into a single 16-channel MIDI device (Yamaha XG MU100, circa 1997), the stereo output recorded untouched. I did the mix, he did the transcription from audio to MIDI.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.