Let's take an accounting approach, based on some
rough assumptions, all offered in the spirit of
good humor or

, and not saying it can't be done.
- The problem of generating midi from an arbitrary mp3 might be somewhere between 3 and 20 times as complicated as Melodyne.
- Melodyne is currently 126 megs of compressed binary.
- The compressed image might be approximately 1/4 executable, with the rest being graphics, symbols and OS overhead requirements.
- Compressed executable code might be compiled at a ratio of approximately 20 to 100 bytes per line of source code, average.
- A great programmer maybe could write somewhere between 10 and 100 lines of code per day, average, taking into account debugging, research, etc.
- Multiply software work by a factor of 3 for other workers for image generation, documentation, testing, infrastructure, etc.
- Number of working days a year, with holidays and a couple weeks of vacation: 240.
- Working MIDI cable, two to three weeks
.
Certainly #5 is debatable, with some saying the best programmers achieve best results by
deleting lines of code, not generating new ones. And maybe the 126000000/4 bytes of binary executable includes libraries previously developed by others; best not to duplicate the work of others. Really, all of the assumptions above are highly flexible, and all offered in play. And there could be some math misteaks.
Calculation of staff years to develop a program to convert an arbitrary mp3 program to MIDI:
- optimistic: 3*3*(126000000/4)/(100*100*240) ~= 118 (roughly equivalent to 900 thousand people working 15' each one time, or 378 people dedicated to working 15' a day for 10 years).
- pessimistic: 20*3*(126000000/4)/(20*10*240) ~= 39400 (roughly equivalent to 300 million people working 15' each one time, or 126000 people working diligently 15' a day for 10 years).
This doesn't take into account the time to wait for some genius to develop some appropriate technology. For example, the time between the introduction of the Fourier Transform and the general availability of the modern Fast Fourier Transform was somewhere around 160 years. The FFT is just one contributor to the internals of this project (though overclocked CPUs might make this factor less important).
I have put in my 15' for the day, and I enjoyed it. Best wishes to you, ramtin, for organizing this project.
Bill B.