next »

[Rich Nagel]

This is something that has haunted me since March of 2005.

After moving to my new locale in 2005, I was working on a customer's laptop, and found a real nice sounding MP3 on it. The music is a blockbuster movie sounding sort of thing (along the likes of works by James Horner and John Williams). I swear the music sounds familiar, but can't for the life of me remember what it's from (or if I've ever actually heard it before). The song is quite long, lasts for 9.5 minutes, and sounds very "movie-ish" to me.

Anyhow, since I'm still on a miserable dialup, I've re-encoded the MP3 to a 11,025/Mono/16,000BR MP3 (the CD quality original was 12 megs), and uploaded it to one of my sites ( http://linkremoved.com/Unknown.mp3 ).

Any idea of what/where this music came from or it's title?

[Dotpitch]

Any idea of what/where this music came from or it's title?

MusicBrainz pointed me into the direction of Reflections of Earth, used by Walt Disney World in the Illuminations fireworks show since 1999 (YouTube). It's on their Millennium Celebration CD.

Gavin Greenaway is the composer for IllumiNations: Reflections of Earth. Greenaway's colleague Hans Zimmer, composer of The Lion King, asked Greenaway to take on the project because he was busy with other projects. Zimmer collaborated with Greenaway in the beginning of the process. (link)

[Rich Nagel]

A MILLION THANKS, Dotpitch!!! This had drove me crazy for quite some time now, and I'm quite happy to FINALLY know the origin of this song

According to the Wikipedia, I'm thinking that I might have originally heard the song on ABC sometime throughout the years (the Wiki mentions that the song was used on several occasions by ABC).

BTW, how did you manage to find the info on this song (MusicBrainz)? Had you heard the song before?

Again, many thanks!

[Dotpitch]

A MILLION THANKS, Dotpitch!!!

You're welcome .

BTW, how did you manage to find the info on this song (MusicBrainz)? Had you heard the song before?

I hadn't heard the song before, but it indeed sounded like a soundtrack. MusicBrainz is somewhat like CDDB, but they've added a lot more information for every track, for example audio fingerprints, which enable you to look up unknown music in their database. There were no automatic matches on the fingerprint of your file (because of the terrible quality), but a manual search showed Reflections of Earth (playtime 9:31, like your file) was the best match. A quick lookup on YouTube confirmed it .

[Rich Nagel]

but a manual search showed Reflections of Earth (playtime 9:31, like your file) was the best match. A quick lookup on YouTube confirmed it .

Ahhh... makes perfect sence now I suposse that there aren't all that many songs that are so long, 'twas a good idea searching for the songs duration.

[Dotpitch]

I suposse that there aren't all that many songs that are so long, 'twas a good idea searching for the songs duration.

Actually, the hit was on the audio fingerprint, the duration was just a confirmation.

[Rich Nagel]

Actually, the hit was on the audio fingerprint, the duration was just a confirmation.

How exactly does this "audio fingerprint" function (I suppose, as a musician and an audiophile I prolly should already know this <G>)? I'm guessing that the software on their server retrieves the MP3 file, scans it for volume lows/peaks and such, and then compares it to data already in their databases?

[Dotpitch]

How exactly does this "audio fingerprint" function? I'm guessing that the software on their server retrieves the MP3 file, scans it for volume lows/peaks and such, and then compares it to data already in their databases?

MusicBrainz works with a client application, which calculates the fingerprint of the first two minutes of the file.

Open Fingerprint Architecture normalizes, extracts frequencies, and creates an array representing the frequency data in an audio file. (link)

This fingerprint is sent to the MusicDNS server, which returns a PUID (Portable Unique IDentifier) if there is a match. This PUID is then looked up in the MusicBrainz database to get all the metadata.

The creation of these PUIDs and the matching algorithm are completely closed source. The algorithm can recognize similar music (for example for the MyDJ-plugins for WMP/Winamp/iTunes, to create genre-playlists), is quite insensitive to encoding and transmission artefacts (which is required with lossy compression and analog radio transmission), but is still able to discern different master recordings.

[Rich Nagel]

Thanks for the detailed info on that, Dotpitch

Sounds like an *extremely* interesting method... I definately need to read up a bit more on this.

With the frequency recognition, that almost sounds to me like a good start for a MIDI->WAV converter (which, I'm sure as you know, isn't too reliable (or not even working at all) these days).

[saga]

i guess you mean WAV->MIDI? There is a big problem: A fingerprint is just a fingerprint. It doesn't need to contain detailed information about different voices, it's just for the complete mix. MIDI->WAV will never be perfect. It if it was possible, we would also have perfect time stretchers and pitch shifters, but it's simply not possible.

[Rich Nagel]

i guess you mean WAV->MIDI?

Yes, that's exactly what I meant... the 'ole brain doesn't seem to engage properly these days <LOL>!

There is a big problem: A fingerprint is just a fingerprint. It doesn't need to contain detailed information about different voices, it's just for the complete mix. MIDI->WAV will never be perfect. It if it was possible, we would also have perfect time stretchers and pitch shifters, but it's simply not possible.

Ahh... makes perfect sense now... (re: "It doesn't need to contain detailed information about different voices, it's just for the complete mix."), which is the biggest problem with all of the exisiting 'so-called' WAV->MIDI converters.

[KaoDome]

@Dotpitch: Hmm... which program do you use for the fingerprint/matching? The only way I use MusicBrainz is for tagging purposes (already knowing the track/album), I didn't know it could be used to try to know info on an unknown track.

Thank you very much.

[Dotpitch]

@Dotpitch: Hmm... which program do you use for the fingerprint/matching? The only way I use MusicBrainz is for tagging purposes (already knowing the track/album), I didn't know it could be used to try to know info on an unknown track.

The Picard tagger automatically looks up metadata based on fingerprints, so not based on the present tags. If there are not automatic matches, you can look at the details for a track, copy the MusicIP PUID and use the manual search to find the best match. Ofcourse, the track has to be in the database, otherwise MusicBrainz can never find it.

[Programmer17]

Wow, some excellent work to track down this song. And a nice piece too! Just for reference, there's a cool website out there where you can do exactly this type of thing - post a song or melody that you dont know the artist of. The community then helps to identify it. Check it out: http://www.watzatsong.com/EN/