In the MP3 HQ scene there's an accepted standard of 192Kbps HQ encoded files being the cd-quality rate. I'm wondering if there's any such standard in the .ogg community? I have good enough ears/speakers to be able to tell the difference between Lame/FhG at 192kbps for MP3 files and my primary purpose for ripping cds is to keep backups available on my computer.

Any suggestions would be appreciated.

Thanks, Seth

Quality 6 is generaly transparent and will give you similar bitrates to your mp3's.

Q6 is *generally* transparent? I must have lousy ears, 'cause Q4 stands up to even the most discerning listening for me.

Hmh, that standard is basically hopelessly outdated. Using fixed bitrates for mp3 should really be a no no these days. At least when it comes to high quality encodings and there are no apparent reasons to use CBR settings.

I recomend taking a song that represents the majority of music that you listen to, and starting at like q4 or so, and see if you can tell the difference (blah, ABX it), and then if its not acceptable, up the rate to q4.10 or q4.20, repeat until you get a setting that is cool with you. I got up to q4.9 and coulnd tell much if any difference in the environment, so I upped it to an even 5 to be safe and have been happy ever since.

I guess the best way is to start at q 4 and move up if through ABXing it's not transparent. The problem with Vorbis is that quality is not linear enough. Sometimes q 4 is transparent on somethings and q 6 is transparent on other things. The quality varies a bit. Hopefully this can be fixed soon.

I'm speaking slightly out of my ass, but IMHO, maybe something like -q 5 is equivalent to the "192 cbr standard" of mp3. The lame VBR "transparent" setting (--preset standard) averages more than 192 kbps, and, being vbr, is better. I think the equivalent to that is probably closer to -q 6, if you use one of the forks such as gt3b2 or aotuv (it's looking like aotuv is going to be the better of the two on the majority of samples, but not all of them). If you're using "stock" vorbis, you probably have to go higher.

I question strongly the notion that -q 4 can be called "transparent"; I think it's more like the equivalent of lame --preset medum or 160kbps maybe. Transparent for a lot of people? Definately. Good enough for many peoples needs? Absolutely. Comparable to --preset standard or mpc --standard or 192k mp3? Based on the ABX success of many people in recent 128k listening tests... Definately not.

ABXing yourself probably is the best route, but it is time consuming, and I understand the value of having a "standard". Also, it's desireable to have encodings that are future-proof - in case you get better at spotting artifacts.

I don't consider Vorbis -q 4 to be sufficient for general HQ purposes because it has a slight but verifiable shrinkage of the stereo image (thanks to the lossy IS algorithm at that quality level). This is my personal complaint with Vorbis at low bitrates. It's fair to say "aoTuV -q 2 sounds really good"...until you encode a source with wide stereo seperation and the encoded Ogg file sounds like fat mono. I wish you could tweak the stereo settings with a command line switch.

We have seen some 5's in the Vorbis listening test results at -q 4.35 on some samples, which typically are problem samples, so that's why I believe it can be transparent on some things but not all.

Hey everyone...
Apart from the -q setting, don't you use something for things like low pass/high pass...etc? (i'm just asking, don't know much about those things)

Generally if you don't know what you are doing with the advanced encoder options, they can affect quality in a negative way. So it's best not to touch those unless you need to and know what they do.

This question is being asked regularly.

Why not add something to the FAQ to say "for 99% of people, Qn is transparent on 99% of music."

I know that transparency varies between listeners, but isn't there a consensus where it is usually transparent, like with MP3 APS and MPC Q5?

Not enough data, I assume.

The 'general' level of transparency is a bitrate which is transparent on 95% of samples for 95% of listeners as far as I can remember. I'm not sure how useful asking people to test this themselves and forward on the results would be, for one thing feedback isn't likely to be high enough, and secondly this community doesn't represent the average listening capability of the general public with people like Guruboolez & Harashin around

as i suggested before, xiph should do a public ABX test at various Q levels to gather data for a 95th percentile q setting.

i really think this needs to be done, it would help ogg acceptance if users could know with confidence that a specific q setting by default will likely be transparent for them.

a large enough dataset of listeners will offset the few 'golden ears' out there.