80 kbps personal listening test (summer 2005)
Reply #81 – 2005-07-17 14:29:23
(...) or maybe calling into question the way recommendations are done. [a href="index.php?act=findpost&pid=312704"][{POST_SNAPBACK}][/a] I'm open to suggestions. In the beginning, we didn't even have a recommended vorbis encoder thread (...) I suggest that we (note that I say "we" and not "you") don't recommend any specific encoder. There are many possible reasons to not recommend a specific encoder, but the less acceptable one is the one linked to the lack of listening tests. We have the chance to have people working hours and hours to improve the quality of an encoder (it's not necessary the case), and we can't tell them "bravo, but sorry we don't have any feedback to recommend your work". If the community can't spent one hour of free time to make a listening test, the community don't have the right to give recommendation based on missing knowledge. It's just a sign of respect and the minimum we can do for the creative and innovative persons working for thousands or even millions people. The recommendation should therefore be limited to explain the use of the encoder: • why users shouldn't tweak the default profiles • how people could solve some issue (pre-echo with --ITP as example) • explain all known problems with the encoder (specific artifacts) • why VBR and not ABR; tips to avoid playback problems with some flash players .... There's no need to recommend an encoder, especially older one over newer and not-tested ones (I don't like the expression "non-tested": it supposes that developers are throwing their work to the public without bother to test it themsleves — and it also supposes that the lack of report is a sign of lack of tests; I'd rather say: no news = good news ). Couldn't we simply precise that: • 1.1.1 is the latest CVS version, that quality is essentially based on the work performed by Aoyumi for aoTuVb2, and that this version is obviously not the most advanced version of vorbis? • aoTuV b4 is the latest version, and that this encoder solves x, y and z issues or bug, and consequently should offer the best results? • there are also other encoders, {slightly} outdated now, like GT2/GT3, Megamix, QK, ModestTuning and give of short description of them ? I know that people like and request explicit answers to their problem: they want THE best encoder and THE transparent setting. As someone starting to have a good experience of multi-codec/multi-format blind listening comparisons, I can say that such encoders and such settings don't and probably won't exist. aoTuV b4, lame 3.97, Nero AAC... could outperform all their competitors, there are still problematic samples for them, some problems audible with the most advanced encoders but not audible with less recommendable encoders.