# Can I use fpcalc to prove two audio files produce the same sound?

I'm using fingerprinting techniques to detect duplicate audio files on my music library.

The first and easy step is to compare the file-size and if they match I use a md5 sum to compare them.

The problem is sometimes the meta-tags of an audio file have been modified. So I compare the length (in seconds) of the file and if they match I use fpcalc to compare them.

When I finally have two audio files with different md5 sums but the same result of fpcalc I use picard (musicbrainz) to normalize the meta-tags (removing any tag that isn't a "standard" tag). Sometimes after saving the audio files they become identical but sometimes they are just different.

• Can I use fpcalc to tell that two files have the same audio with the same quality?
• If not, are there any other tools to compare audio files?

I made a small script to make all this easier https://github.com/falkartis/duplicates

If you think I'm doing it wrong can you please tell me.

In my collection I have lots of different file-types like wav, flac, ogg, mp3, wma...

• What is fpcalc? – MBaz Jun 7 '16 at 21:54
• It's a command line tool to make an acoustic fingerprint of an audio file. Musicbrainz/Picard uses it to fingerprint audio files and lookup the correct tags for the track. It's like shazam or soundhound. Should I add this to my question? – Falk Jun 7 '16 at 23:02
• yes, you should. You see, this website is a bit about signal processing as in "we're the ones that do the signal processing library, not the ones that use that library"; so maybe we can help you, but every bit of knowledge about the algorithm employed by fpcalc would surely help. I don't think anyone of us has used it, and frankly looking up what algorithms it uses so we can answer your question might be a bit much to ask. – Marcus Müller Jun 7 '16 at 23:16
• I see, maybe I'm asking at the wrong place. I tried to understand the algorithm myself, but I'm not smart enough (yet) that's why I asked. – Falk Jun 8 '16 at 9:12