Similarity Forum

General Category => General => Topic started by: WimYogya on July 01, 2016, 10:26:33

Title: Duration inconsistencies
Post by: WimYogya on July 01, 2016, 10:26:33
When comparing Similarity often finds duplicates with a wrong (?) duration.
1. Abba - Waterloo - 4.17
2. Abba - Waterloo - 3.45

When you play/test both, they sound the same. Often the longer version is just a few seconds +/- than the shorter one. The Similarity player box (in this example) has a width for 4.17, but the audiospectrum and the sound stop at 3.45 - suggesting there is a lot of silence after the track. But that is not the case! When I play this 'long' version in my player (MusicBee) there is no extra silence at all.

How come the Similarity duration count (apparently?) seems incorrect? Or does MusicBee make the mistake and IS there indeed extra silence?
It makes it quite difficult to check all the duplicate pairs where only the duration is different...
Title: Re: Duration inconsistencies
Post by: Admin on July 02, 2016, 14:47:23
Some decoders can't give exact duration of audio file, ang give some approximation.Can you send this problematic file to our email to analyse.
Title: Re: Duration inconsistencies
Post by: WimYogya on July 02, 2016, 15:06:28
I wish it was ONE file, there are many, hundreds...
At the moment I am ready to go on vacation with no pc access.
Will send you examples after I am back.

I just found time to send you a screenshot of one of the examples. See attachment.
Based upon my settings Similarity found 2 (almost identical) versions of Only You, by the Flying Pickets. It SAYS one file is 3:55, the other 4:13.
But in my player I have 3 files, two of them 3:23, one of 3:25. Very different from what Similarity found.
And with the Similarity player the audiospectrum shows 3:23 within a total track length of 3:55 - which includes: a dead end.
After I sent the screenshot I played the second version in Similarity, with similar result: audiospectrum 3:25, given duriation 4:13
So, what to think of that and how to get more precise results?

Apart from this: is there any program that could DELETE unusually long silences from tracks?
I only know that I can instruct my player (MusicBee) to decrease silences to (for instance) 2 seconds, but the actual LENGTH of the track remains.
In Audacity I can remove the silences, but that is an awful lot of work.
Any better tool?
Title: Re: Duration inconsistencies
Post by: Admin on July 09, 2016, 23:40:03
Please, send this file to us, I think the problem with VBR coding method.
Title: Re: Duration inconsistencies
Post by: hsei on November 23, 2016, 22:15:09
A good tool for editing MP3s (e.g. removing silence) without recoding is MP3DirectCut.