No, I meant Ail phone



I got an Iphone at the place I work, so I was able to test various things.
One that could be interesting for some readers is Shazam, an App that automatically detects the song you're listening to, giving you the band name, song name and album cover...
It's a quite successful App (Software) and peoples seem to enjoy it because the songs happens to be detected in few seconds, and it seems to be working quite often... So I decided to make a few tests to see how far it could go.

Let's recognize some more or less underground records...

To begin with, I choose mostly bands to be quite famous:
Napalm Death: "It's a M.A.N.S World" (Enslavement To Obliteration CD).
    -> Worked
Bathory: "Hades" (Bathory CD).
    -> Worked 
Exhumed: "Excreting innards" (Chords of chaos split CD, later re-released on a CD compilation).
    -> Worked
Exit 13: "Societally provoked genocidal contemplation" (Ethos Musick CD).
     -> Worked

Since these bands were all released on quite big labels, let's try something more obscure, or released on labels that stopped their activities:

Nausea: "Mind dead" (Crimes against humanity CD).
    -> Didn't work
Dead Infection: "From the anatomical depths" (Chapter Of Accidents CD).
    -> Didn't work
Sacramentum: "Fog's kiss" (Far away from the sun CD).
    -> Didn't work
Corpus Rottus: "Vomit pool" (Rituals Of Silence Lp).
    -> Didn't work

So this will work if someone from a label put the songs on Itunes, legally or not... (You won't necessarily find something that was released 15 years ago on a very small obscure label that did only two records... But the software detected songs from Viogression, so who knows...)


Let's change the sound of the music.

I took a song from a later Napalm Death record that was detected easily in "regular sounding" conditions.
Then I decided to delete a good part of the mid-range and high-range frequencies with an equalizer (This made the sound very deaf, with much more bass frequencies than the original).
    -> It worked.

Then I decided to make the game a little harder, and pushed the bass frequencies much higher. The result in sound was quite overdriven bass drums, and it began to be noisy.
    -> Surprisingly it worked.

I thought it wasn't cool, so I pushed the bass frequencies higher again, it became more overdriven, sounded more like harsh noise (Even if a human being could hear Napalm Death behind)...
    -> And hopefully it didn't work :)

Let's change the speed of the song.

Then I decided to decrease the speed of the song.
To make the process a bit quicker (My computer is a bit slow sometimes)
 I took only 30 seconds from a song (Exactly 32.719 seconds).
The first step was to make the song a little slower, the extract became 33.046 seconds long.
    -> Detection worked.
It also worked for the next step (33.373 seconds long), but not for the following one (33.701seconds long).

Increasing the speed of the song brought almost the same results, it was just a little little better.
So the software will find what song it is if you don't play him something too "damaged"...

Includes Usb 3.0 port.
         Compatible with Windows 7.

Let's try to detect an unofficial live recording.

This could be nice to try with an unofficial recording of a song that was already in the database, to see if the software can detect this song.

First I took two bootleg grind songs, but it worked for none:
Napalm Death: "The Kill" (Live in Europe 1998 - Scum dead corporations bootleg)
Napalm Death: "Prison without walls" (Live in Europe 1998 - Scum dead corporations bootleg)

I thought it might be a bit too complicated for the software to detect the same grindcore song but in other conditions, since the style can be chaotic.

So I decided to make it with easier listening music, and choose "Paranoid" from Black Sabbath (Which might be one of their more "radiofriendly" tracks).
Let's go on Youtube, and try the official recording from this song, to make sure there's no problem with the website for a reason or another: Shazam detected the song from the official video, no surprise since it's the same recording than the album, but it worked despite the quite bad quality sound (Youtube mp3 encoding) so we know Youtube is ok for detecting.

Then I tried with various live recordings:
Black Sabbath - Paranoid (Live in Paris 1970).
    -> Didn't work, despite of the quite "clear" sound
Black Sabbath - Paranoid (Live at Worcester 1983).
    -> Didn't work
Black Sabbath - Paranoid (Live in Philadelphia 1985).
    -> Didn't work

So Shazam doesn't detect the song, but the audio recording. You could give him 30 different recordings of the same song but he wouldn't know it, on the other hand you could give him 10 versions of the same recording (With a "special effect" such as more echo, chorus, more bass frequencies) and it could work.
This software could seem like it's super-intelligent, while it is not: It only compares one audio recording to others in the database. This said, I don't think we could claim it's not intelligent since my tests (Deleting high and mid frequencies) shows it's not as dumb as A = B and probably compares the recordings under various angles (Amount of peaks, shape of the audio spectrum...)

I imagine you could record the noise of your water closets, then register it on Itunes, then play the same recording to Shazam, and he would detect it's the same recording... But unfortunately, if you play him another recording, of the same water closets making the same noise, it won't work.

Something like CONCLUSION:

If you're searching for very precises songs (Very underground recordings that are not in the database, unofficial recordings...) or tracks from a damaged support (A tape/ vynil that plays a bit accelerated or "strange") you won't have good results. But if you think about probabilities, I imagine most of users will probably request songs released the last years (Or quite famous ones), or something they heard on Tv, so a good part of requests will have good resulsts.

Now, if you can afford an "Hail phone", if you agree to create an account giving your bank account number, this could be a nice software to test.