Strange duplicates

[gah789]

I have been experimenting with the unregistered version of Perfect Tunes. The De-Dup option seems to be a bit erratic. Some of the duplicates which it identifies are genuine but for example I have a number of different recordings of classic piano pieces such as Beethoven's Diabelli Variations. When comparing FLAC files, De-Dup manages to convince itself that the same variations from discs by Schnabel, Brendel & Lewis are duplicates. Similarly, it identifies movements of Beethoven Piano Concertos played by different orchestras and soloists as duplicates. While they are very similar in length, I find it hard to believe that they are identical to one frame. All of the CDs were ripped using dbPoweramp CD-Ripper and were checked against the AccurateRip database when created. The pieces are not identical.

Another oddity is that it believes that all of the tracks from a recording of Bartok's Miraculous Mandarin are duplicates of each other. This appears to be a misinterpretation of the original source information. This disk was one of a very small number of Naxos DVD-Audio disks that I purchased and converted to WAV files and then to FLAC.

[Spoon]

Re: Strange duplicates

DeDup is based on audio finger printing, that is if there is and original track and a cover version of the track, they will be indicated as duplicates because the underlying audio is similar.

[gah789]

Re: Strange duplicates

If the audio finger printing can't pick up the differences between different performances of concertos or works for solo piano, then we may assume that it will be of limited use to anyone with a large classical music collection. No musician would regard them as being equivalent for any interesting purpose. This is a lossless format so we are not dealing with low quality MP3 files.

[BobTheDan]

Re: Strange duplicates

While I was browsing my music library I found several music files with duplicates which have been not detected by Dedup, but certainly it should had detected based on my experience with Dedup since it's beta release. Some of the music are just the same with few seconds difference of their length. But some may have small differences large enough for Dedup not to detect?? Just thought may be it would help you to improve the algorithm of fingerprinting if I submit those files to you.. Let me know if you think it's worth to have a look at them.

[Spoon]

Re: Strange duplicates

A large time shift (over 5 seconds) would affect de-dup.

[BobTheDan]

Re: Strange duplicates

Except for one track, rest of the five's music starts playing almost at the same time frame from music starts playing. One track is only one second off from it's duplicate and it sounds very same to the counterpart. The rest's total length is off up to 5-6 seconds. So unless Dedup take the fingerprint from the track's last 5 seconds (from the way I understand how it works..) it should detect those files and it has done it in the past for many different music files, that could sound very different to the ear, but was the same song for an example like tracks from different audio CDS with different masterings..etc.