Right, I've attached a FLAC of a 20-second sample of Annet Louisan's "Das Spiel".
In another topic I stated that I was able to ABX this and a q7 Vorbis encode of it, so Phong kindly asked I upload this as he suspects it might be a problem with Vorbis itself.
I do not know for myself.
The instruments at q7 are perfect to me anyway, the way I could tell was the voice. No artefacts that I could hear, but when listening to both back-to-back you might be able to catch subtle differences in the way the singer sounds.
It's difficult to explain (don't think I have the jargon for this), but it sounds less... pure? or less alive, rather.
That's the main reason I was able to ABX them from one another.
Another (even less important) point was that the Vorbis encode sometimes lost some of the noise that made it a human voice, like how in the source you can hear the tongue movements, but the Vorbis loses some of them.
My encoding was done with libVorbis 1.1.0 and the aotuv b3 patch, and I suspect it might have to do with that??
EDIT: The ABX was done with Foobar2000's plugin.
EDIT #2: I just checked again. At the end of the 2nd second/start of the 3rd one, the word "warst" in "was Du einmal warst" (pronouned sthg like "vas doo I'n mal waars") sounds different. In my Vorbis encode it sounds like she's just eaten joghurt (sp? yoghurt?)
Well, I was pretty interesting in sample you can ABX at Vorbis Q7, so I tried at Q4 (AoTuVb3 original compile) and I admit I did'n recognize any difference from original. So maybe I'm totally deaf but anyway it seems odd to me
Just for comparsion I've uploaded two samples which I use for testing encoders. It's impossible for me ABX the first one at Q4 (=transparent), but easy with Q3, so that's why I use Q4 (I use many others too ) With lame (3.96.1) V4 still easy to ABX, with V2 it's transparent.
So if you or others are interested in testing it too, I would be very interested in yours results / opinions.
Cheers, K.
[attachment=1387:attachment]
[attachment=1388:attachment]
Well, I was pretty interesting in sample you can ABX at Vorbis Q7, so I tried at Q4 (AoTuVb3 original compile) and I admit I did'n recognize any difference from original. So maybe I'm totally deaf but anyway it seems odd to me
Just for comparsion I've uploaded two samples which I use for testing encoders. It's impossible for me ABX the first one at Q4 (=transparent), but easy with Q3, so that's why I use Q4 (I use many others too ) With lame (3.96.1) V4 still easy to ABX, with V2 it's transparent.
So if you or others are interested in testing it too, I would be very interested in yours results / opinions.
Cheers, K.
[attachment=1387:attachment]
[attachment=1388:attachment]
[a href="index.php?act=findpost&pid=281128"][{POST_SNAPBACK}][/a]
Can't blame you for not telling the difference... maybe I heard this artist far too many times
I'm not very good when it comes to instrumental sounds unless there is an artefact, but I tried your first sample. Couldn't ABX it at Q4, but could at Q3, just like you.
Don't have time to try with lame, I am at work after all...