what if we choose an ogg vorbis quality standard which is transparent? i mean what most people think its transparent, and set the minimum quality. like mpc, starting from quality 5, lame from --alt-preset standard.
if somewhat this post offended anybody, sorry im a noob

For me, aoTuVb2 at -q 6.22 was transparent on those 11/18samples.
Vorbis_q6.png

I love how it's 6.22 and not say... 6.21 or 6.23

Which means that for most people that around -q5 or -q5.5 will be transparent

Bearing in mind that -q4 scored 4.59 out of 5 overall in the recent 128kbps listening test as well, I would be surprised if many people could ABX -q 5.

P.S. I'm referring to the AoTuV beta 2 version of course.

thanx ya all. a long time ago, i use the gtb2 encoder to compress my audio cds at q6. and if you tell me to distinguish between a q2 and the original wave file, seriously i can't even if i use my phillips hp890 headphones or ada305 speakers. but when i use defx (for karaoke stuff) for winamp, i can hear watery effect on a 128kbps mp3, or q2 ogg vorbis. but the thing is i used mpc now, and the watery effect still can be heard on the standard quality. but using ogg vorbis q6, the watery effect is none.

hmm, now im thinking changing back to ogg vorbis. ok thanx for your replies.

Yes, harashin could you tell us why 6.22?

off topic anecdote:
I have a set of pro-logic speakers somewhere. It's fine for CD music, as reverbs and stuff reflected to the rear speaker and it helped make the music sound "fuller" (without sounding artificial), but when I played a 128kbps MP3s, what went to the rear speaker was nothing but bubbly artifacts.

To duplicate the same effect, just invert one stereo channel against the other.

I tried to compare aoTuV with GT3b2 -q6 at (nearly) same bitrate.
http://www.hydrogenaudio.org/forums/index....ndpost&p=215109

Personally I went through and couldn't distinguish anything above 4.85 or there abouts, so I rounded to an even 5 when I do my encodings (using the stock 1.0.1+). Granted, I'm doing mine in mono, cause I can tell a difference up till 6 with the stock encoders in stereo, but anyway my point:

everyone's ears will be different, Xiph picked q4 as the entry/average bitrate people *might* want to use, but doesnt overly promote it. My father cant tell anything is different (again in mono) above q4.1, while I can above (just bearly) 4.8, and using ER-6's I can get it up to q5.2. My girlfriend peaked out at 4.75 in her tests. But yeah, everyone's different, some can tell a difference, some cant, and as was put to me when i first read about ogg vorbis, if you cant tell a difference, you dont need the extra bits unless your doing an icecast and can do bitpealing later since in theory, you cant hear the difference anyway.

So a transparent standard really doesnt exist. It might for a segment of users, as an average, but thats as close as it will get.

my 2 cents.

-- Page

You may want to look at this thread for why that peculiar -q setting was chosen.

As it is right now, I fear that vorbis doesn't have a -q setting that is a direct analog to --preset standard or mpc --standard. In both of those cases, it was decided that "this is the setting that will be deemed transparent" and any samples that weren't at that setting were/are considered bugs or at least specific targets for improvement. The quality settings for vorbis are more vague... -q 5 is better quality than -q 4, and -q 6 is better than -q 5, but that's about it.

It seems like vorbis quality may be somewhat uneven... There are a number of samples that are not transparent at -q 6 or even higher. Likewise, I think there are probably a lot that are transparent below -q 5. This is also true of lame and mpc, but I think to a lesser degree. The confusion is only made worse because there are now a number of forks and it's not clear that any of them is better all the time - some are better on some samples, others are better elsewhere.

It would be nice if vorbis developers picked a -q level that would be defined as the target for transparency. The bitrates for -q 6 are a little lower than lame --preset standard, and a fair amount higher than mpc --standard, so I would hope such a target would be lower than that, but the number of problem samples at that level is disconcerting. Then again, I may be overreacting.

I'm thinking a variety of test samples should be selected, and then an ABX test should be done with as many listeners as possible, and 90th percentile chosen for transparent Q setting.

Actually the mere fact that I can adjust the quality with some granuality is the reason I started to fancy Vorbis (plus it is gapless and better tagging than mp3) I keep all my music in FLAC for archive, but to listen in various PCs and later portable I want lossy. However I do not think I have golden ears so I plan to do some ABX with some of the test samples this community gathered and set a quality level that is good enough for me. Good granuality makes the whole thing more tailormakable if you like.

Triza