Codegen delivers different Fingerprint for 2 almost identical Songs

[mhuhn]

Hello,


i just want to try out this new toy.
At last i want to filter oput dupes in my collection.
I have implented a binary only Solution which is working with MD5 -
which works fine.
Naturally it fails when i have 2 identical Songs but they have a
different Bitrate.
So codegen comes in my mind.
I used codegen with the following parameters:
codegen Odd_1.mp3 10 30 > compare1.txt
codegen Odd_2.mp3 10 30 > compare2.txt


Technical Info of the "Odd_1.mp3":
Payload-Größe: 3261137 Byte
Header gefunden bei: 22476 Byte
Länge: 204 Sekunden
MPEG-1 Layer 3
128 Kbps, ungefähr 7802 Frames
44100 Hz Joint Stereo

CRC: Nein, Copyright: Nein
Original: Nein, Hervorhebung: Keine

Technical Info of the Odd_2.mp3:
Payload-Größe: 3286183 Byte
Header gefunden bei: 512 Byte
Länge: 205 Sekunden
MPEG-1 Layer 3
128 Kbps, ungefähr 7862 Frames
44100 Hz Stereo

CRC: Ja, Copyright: Nein
Original: Ja, Hervorhebung: Keine

So the only difference is "Joint Stereo" vs "Stereo"

Now i compare the 2 result files and the information in the field
"code" differs.
I used codegen V4.12 under Windows 7 x64

Have i missunderstood the meaning of codegen?
Shouldn't i get 2 identical codes? (Audio fingerprints)?


Please advice me!

Best regards
Martin

[Alastair Porter]

Hi Martin,

If the two files you are testing differ, then there is a chance that the generated codes may be slightly different, as you have found out.

In order to determine if the two songs are the same, you will need to do some further work decoding and comparing the generated codes. Take a look at this post for an example of doing this:

http://musicmachinery.com/2011/06/25/finding-duplicate-songs-in-your-music-collection-with-echoprint/

Alastair

[Martin Huhndorf]

Helo Alastair,

perfect!
Now nearly everything is clear.
Except: PLEASE make ist easier to have my own echoprint Server! The requirements for this thing is outstanding :)

Best regards
Martin

2011/8/1 Alastair Porter <alas...@echonest.com>

[Manan Vyas]

Hi Alastair,

I am also trying to achieve similar thing - basically trying to check if given 2 audio clips - check if one is subset of another or not, I checked the link you shared and I feel I can probably use echoprint/echonest for the same, just one question - I am not clear on the part about how to ingest and index my audio files, I checked the API doc shared on developer.echonet site but nothing much useful. Does the echoprint server support such functionality, if yes, do you know how? It will be extremely useful if you can share some details on the same. Thanks in advance.

~Manan