Description

Have you got mixed DJ music lying around? Wish you knew what tracks were in the mixes? Wish a long 3 hour set was broken down into separate MP3s by track? I think this is doable.

Technologies

I believe we can use MusicIP's MusicDNS and libofa to do automated analysis of DJ mixes. I think this can be done using a simple scripting language. I prefer Ruby, but Python, Perl, or even shell scripts should work.

Tasks

I think we can breakdown the project this way:

• Part 1: Write a wrapper around libofa in scripting language of choice (Python wrapper may exist)
• Part 2: Write a simple script that takes a DJ mix MP3 and copies out a 2-minute time slice every 5 seconds
• Part 3: Run audio analysis on the 2-minute time slices in parallel
• Part 4: Submit the audio analysis to MusicDNS to look up
• Part 5: Generate a list of all of the occurrences of each track in the mix including 'unknown'
• Part 6: Determine a simple algorithm (probably first occurrence or first two sequential occurrences) to figure out where in the list a given track 'starts'
• Part 7: Split original MP3 DJ mix according to the list data in Part 6
• Part 8: Edit ID3 tags in split MP3s and add any relevant metadata

Tools

TBD