fdmf 0.0.9r

fdmf is portable perl/C software for finding pairs of music files in a collection that are likely to contain the same music. It works on the music itself, not on the filename, tags, or headers. The project uses an audio fingerprint, or perceptual hash to recognize the duplicate files. It is currently under heavy development, so it might be buggy, broken, or otherwise bad. But it works for me. Please email me and tell me if it works for you. Bug reports are appreciated.

It would help the development of this software if I had a larger set of files to test it on. If you feel like lending music files to this effort, it would be greatly appreciated. The optimal form would be a CD or DVD of MP3 or OGG files which could be sent via postal mail. The media will be mailed back to you after it has been used, along with some extra disks of music for your testing purposes. Please email me if you are interested in lending files.

USAGE:

fdmf < music_dir >
vector_pairs

USAGE EXAMPLE:

./fdmf /home/bob/music/
./vector_pairs

vector_pairs will print on stdout a list of pairs of files that may contain the same music.

NOTE: The first time you run this program it will do a lot of processing. It takes 10 minutes to process a music_dir containing 140 mp3 files on my Apple G4. It will attempt to cache the result so that subsequent runs are faster. You can interrupt it and your progress will not be lost.

ANOTHER NOTE: On my current development box, an IBM T20, mplayer is the fastest mp3 decoder. mpg123 is also quite fast, much faster than mpg321. Some Linux distros come with mpg123 being a symbolic link to mpg321 since mpg123 is not under GPL. Bottom line: you can use any program for decoding if it can take the name of a music file as a command line arg and write the raw decoded 44100/stereo/16-bit stream on stdout. You might want to take some measurements on your system to see which decoder is fastest.