Ogg -q 0 comments Options

[ff123]

I tried several samples using the latest ogg compile I could find (7-4-02 on rarewares) and tested using -q 0. For reference, I also compared with mp3pro at 64 kbit/s (MMJB 7.2) and wma8 at 64 kbit/s.

Here's what I heard:
charlies, ogg bitrate = 63 kbit/s

CODE

1R = charliescharlies.wav

2L = charliescharlies_mp3pro.wav

3L = charliescharlies_wma.wav

4R = charliescharlies_ogg.wav



---------------------------------------

2L File: charliescharlies_mp3pro.wav

2L Rating: 1.8

2L Comment: flanging

---------------------------------------

3L File: charliescharlies_wma.wav

3L Rating: 3.4

3L Comment: metalling ringing sounds

---------------------------------------

4R File: charliescharlies_ogg.wav

4R Rating: 2.1

4R Comment: stereo collapse, plus a blurred sound, which makes the singer sound like

he's wheezing.

---------------------------------------

wait, ogg bitrate = 65 kbit/s

CODE

1R = waitwait.wav

2L = waitwait_mp3pro.wav

3R = waitwait_wma.wav

4R = waitwait_ogg.wav



---------------------------------------

2L File: waitwait_mp3pro.wav

2L Rating: 3.7

2L Comment: ride cymbals strikes are mushy

---------------------------------------

3R File: waitwait_wma.wav

3R Rating: 3.1

3R Comment: ride cymbals are warped sounding, and the image shifts around

---------------------------------------

4R File: waitwait_ogg.wav

4R Rating: 2.1

4R Comment: severe stereo collapse

---------------------------------------

deerhunter, ogg bitrate = 38 kbit/s

CODE

1R = deerhunterdeerhunter_wma.wav

2R = deerhunterdeerhunter_ogg.wav

3R = deerhunterdeerhunter.wav

4L = deerhunterdeerhunter_mp3pro.wav



---------------------------------------

1R File: deerhunterdeerhunter_wma.wav

1R Rating: 3.3

1R Comment: occasional smearing of guitar plucks; some noise pumping

---------------------------------------

2R File: deerhunterdeerhunter_ogg.wav

2R Rating: 1.3

2R Comment: yuck!  guitar has a fuzzy sound; lots of transient smearing

---------------------------------------

4L File: deerhunterdeerhunter_mp3pro.wav

4L Rating: 2.6

4L Comment: mushy guitar plucks; some flanging

---------------------------------------

ff123

[Caleb]

so what ur basically saying is that the current ogg vorbis is crap heh.

and WMA is #1.

edit: i didnt notice that he was talking about -q 0 - im blind as shit

[elfin]

It is not crap at all.

[HotshotGG]

yuck!  guitar has a fuzzy sound; lots of transient smearing

Monty said he would be adding impulse short block tuning and lowpass filtering "switches" to the Vorbisenc API libraries in which he has already accomplished. Those "internals" could find there way into the next release of the physcoacoustics algorithm. Hence you will then have the ability to use long and impulse short blocks for experimenting from my understanding between 64 and 2048 samples. This may come in handy for isolating those transients that would cause pre-echo otherwise. That is not the only possibility theortically, however it works for the time being.

[Caleb]

Originally posted by elfin
It is not crap at all.

i know
i use it and i like it.

but thats what he says..
ow wait he was talking about '-q 0', bah.. who uses it anyway, i mean, maybe is good for streaming or something, i wouldnt think that anybody would encode with that quality (why would u want to create really good rips with EAC and then encode then at -q 0.. no sense in that)

[unplugged]

, -q2

[Jon Ingram]

ow wait he was talking about '-q 0', bah.. who uses it anyway, i mean, maybe is good for streaming or something, i wouldnt think that anybody would encode with that quality

People rip to 64 kbit/s WMA, so you'd be surprised the number of people that would use '-q 0'. There is even a thread on the vorbis mailing list at the moment complaining that '-q 0' is too high quality, and the '0' level should be given to the experimental 48 kbit/s '-q -1' setting which is in CVS.

One aim of Vorbis is to beat MP3pro and WMA8 at 64kbits. With proper tuning, there is no reason why this shouldn't happen. Tuning is the problem. Monty has such good hearing that *everything* at ~64kbit/s sounds awful -- he needs some of us deaf people to help him out . I anticipate that, as the specs are written, and version 1 finally appears, a lot more people are going to take a look at Vorbis, and tinker around with tuning (particularly now that some of the internal parameters have been exposed to oggenc, as hotshot mentioned). There is a lot of scope for improvement -- this is a *good* thing.

Of course, you should note that this is only the results of a test from 1 person. It's possible that he's particularly sensitive to the defects in Vorbis '-q 0', compared to the defects in WMA8@64 (this is something which I share with him -- WMA8@64 sounds very good to me for the bitrate). Hopefully when RC4 is official we'll get a group test set up at some different bitrates ('low' ~ 64, 'medium' ~ 128, 'high' ~ 192) -- codecs that are world beaters at 64kbit/s might not be the best at 192kbit/s.

[elfin]

Actually I use -q 9.2 , but -q 0 sounds great for <64kbps

[Corsair]

Elfin, maybe you should test RC2 Garf Tuned 2 (-b350). That's what I use. I did some tests and found out that I can still hear some pre-echo at -q9.0 (RC3), but I couldn't with RC2 GT2. Also, RC2 GT2 has lower average bitrate (some 20-30kbps) then -q9.0, but it will give more bits to transients then RC3. I mean, pre-echo is the only artifact you're going to encounter at these bitrates (>300kbps).
You can get win32 compiles and source of RC2 GT2 here:
http://sjeng.sourceforge.net/ftp/vorbis/

[Caleb]

Originally posted by elfin
Actually I use -q 9.2 , but -q 0 sounds great for <64kbps

actually thats tooooooo mcuh.

i'd use 9.05 at max, although i mostly use 7-8.

[mithrandir]

Originally posted by unplugged
, -q2

I agree... -q 2 sounds pretty damn good for a nominal bitrate of 96kbps (average bitrate typically is 80-95kbps). There's some stereo collapse but there's practically no annoying frequency artifacts with "normal" music. If I'm not paying attention to the details (i.e. casual listening), -q 2 sounds transparent to me.

[mithrandir]

Originally posted by Jon Ingram
Of course, you should note that this is only the results of a test from 1 person. It's possible that he's particularly sensitive to the defects in Vorbis '-q 0', compared to the defects in WMA8@64 (this is something which I share with him -- WMA8@64 sounds very good to me for the bitrate). Hopefully when RC4 is official we'll get a group test set up at some different bitrates ('low' ~ 64, 'medium' ~ 128, 'high' ~ 192) -- codecs that are world beaters at 64kbit/s might not be the best at 192kbit/s.

I've found -q 0 to surpass other competitors (MP3Pro, WMA8) by a substantial margin. I have not done any ABX tests with problem samples so perhaps Vorbis will come up short there. And granted, -q 0 doesn't necessarily sound good, but I subconsciously forgive less-than-perfect output at 64kbps...we're talking about keeping only 4.5% of the original signal.

I can see why some think -q 0 is too high quality for the lowest setting. My personal complaint is having to use -q 5 to achieve lossless stereo coupling. By the time I increment the slider to -q 3 and -q 4, Vorbis is getting very transparent to me (frequency-wise) so jumping to -q 5 seems like a waste of bits. At that level, I'm ready to use mpc --standard.

[Cygnus X1]

I was a bit surprised by the results of the ABX testing o these codecs at 64kbps. Personally, I know for a fact that I prefer Ogg over WMA or mp3pro at this bitrate. Of course, none of them sound particularly good, but as others have pointed out, there's really not much data left at this ultra-low bitrate to play with. WMA is the most perceptually annoying codec I have ever tested--at 64kbps it is completely unbearable, with strange robotic noises, splats, and metallic-sounding transients and vocals. Mp3pro sounds like how I would imagine picking up a radio station on metal dental braces would sound, i.e. very artificial and metallic in the high-end. For me, the least salient (noticable) low-bitrate codec is Ogg--no horrible artifacts, no completely collapsed stereo image, and no fake SBR highs. Of course, I won't actually admit to using Ogg for listening at -q0 (-q5 is fine for me). I would go so far as to say I would rather listen to an mp3 encoded at 64kbps, 22Khz than WMA or Mp3pro (44.1Khz) at the same bitrate--the mp3 will sound dull, but more enjoyable in that there will not be an overabundance of artifacts, etc. For me, that is the most important consideration of low-bitrate listening. For that, Ogg wins hands down.

[ff123]

A group test at 64 kbit/s when rc4 is released should be very interesting, with all the different opinions on what sounds worst at this bitrate.

ff123

[redcane]

Obviously your ears are different, but I haven't found any tracks that really stand out as problematic on the CVS ogg -q 0, but I can generally pick 96k WMA... I have no experiecne with MP3pro.

I think what I find is there may or may not be more artifacts with Ogg, but they sound so much nicer to my ears. They just really don't stand out as a problem.

[Neo Neko]

Yes the robotic chirps of WMA drive me insane. I can stand a bit of sterio colapse over obvious artificial artifacts. MP3-PRO? I won't touch it. Way to many problems for support and licensing. And of the three it is the worst in my eyes because most of the results are faked due to SBR! SBR can make it sound better. But it is in no way the same song to me. RC4 is a very compelling reason for me to stick with Vorbis. Less robots and less faking.

[jgalt]

I really shouldn't post in this thread as I'm not a knowledgeable person. However, someone asked "Who uses Q-0 anyway." I do. I do not deal with contemporary music. I restore old phongraph record music. Some date back to the late 1800's. Most are mono. I feel very fortunate if I'm able to restore noise free frequency response between 150Hz and 11.025KHz. I have "splurged" at times and used Q 2, or even Q 4.95 but very seldom.

[verloren]

To add an extra wrinkle, I've found that if you measure the file sizes produced -q0 consistently produces files slightly smaller than wma 64kbps. A more direct comparison based on my figures would be -q0.11 which isn't a lot, but does give ogg a little extra boost.

See http://audio.ciara.us/compare.htm for more details (which reminds me, I need to update it with some more results!)

Cheers, Paul

[Gambit]

I used ogg q0 to encode directors commentary from DVD movies. But q0 is not good when encoding quiet sounds. Try to encode Tomb Raider at q0. The background artifacts are too annoying. Actually mp3 doesn't have this problem, but there are other artifacts with mp3.
So I am using q1 now.

[Garf]

Wow, you guys scared me. I started reading from the top and had no idea this was pre 1.0 stuff

--
GCP

[Gambit]

Wow, you guys scared me. I started reading from the top and had no idea this was pre 1.0 stuff

--
GCP

Well I was actualy talking about 1.0 but I don't know why I posted it to an old thread...

[Garf]

I grabbed the clip, thanks

--
GCP