This morning, my friend Charles pointed me to a song on Tumblr that blew up, a remix of Mariah Carey’s “All I Want for Christmas Is You.”
The original poster deleted it, so I’m mirroring it here.
It’s a terrible and amazing thing to listen to — a conversion of the original MP3 to MIDI, and back again to MP3. The resulting version sounds like Mariah as a player piano — none of the original recording is preserved, only a series of hyperactive notes matching the frequencies of the original song.
Incredibly, you can still make out the lyrics and music, though likely only if you’re familiar with the original song.
It reminds me of this German project from 2009, in which Austrian composer Peter Ablinger used a computer-controlled piano to play a child’s voice.
So I had to try out a couple other songs to see how it sounds. I’m pretty happy with the results. Enjoy.
This scene always gets me. pic.twitter.com/u4WcbIZBTM
— Andy Baio (@waxpancake) December 16, 2015
And a bunch more on my Soundcloud playlist.
I used a web-based converter for the MP3 to MIDI conversion, and I used timidity and ffmpeg for the conversion back to MP3.
And, of course, just as I was finishing these videos, my friend Tim pointed out that Tim Troppoli tried this same technique earlier this year. And, go figure, he used Smash Mouth’s “All Star” too. It’s like the Lenna photo for audio.
More great examples in his video, including Piano Man and Stayin’ Alive. I’m not familiar with the Pokemon theme, and it was totally nonsensical to me. I can’t make out a single lyric, a great example of how your brain’s filling in the gaps.
Update: Here’s an example from 2009 of the same technique used on David Lee Roth’s isolated vocals from Van Halen’s Running with the Devil. And the vocals on this version of Oasis’ Wonderwall are particularly clear for me.
the midi to mp3 can be done free online at zamzar.com
what tool would i need to _simplify_ the midi representation, ie making a lower-resolution version by stripping out the less-important notes? ideally i want to turn speech into one note per syllable… is there an easier way to automate this?
I LOVE THIS!!!
I think you need to publish the sheet music for these.
belated realisation– i can speed up the mp3 before converting to midi, then slow down the midi for ‘coarser’ resolution
I LOVE THIS!!!
Comments are closed.