vorbis v1.0

[DSPguru]

vorbis v1.0 is out :
http://www.xiph.org/archives/vorbis-dev/200207/0054.html

[canadian_fbi]

testing it out now on some music (orbital)... yowza. mp3 is dead; long live vorbis :)

[fiorettoe]

Where can we download it?

[canadian_fbi]

i got mine here: http://www.saunalahti.fi/~cse/oggenc.zip

[Defiler]

Grab it out of CVS and compile it, or wait for it to show up here:
http://www.inf.ufpr.br/~rja00/ogg.html

[fiorettoe]

But Vorbis 1.0 is RC4? :confused: :confused:

[Defiler]

It looks like they'll be skipping over RC4. Someone just asked that question on the mailing list, but no one has replied to it yet.

[ChristianHJW]

Finally !!!

Long live opensource !

Yesterday night the CVS wouldnt compile, because Monty had been uploading an incomplete/damaged file ... and i cant test it until i will be at home in a couple of hours !

[raistlin2k]

nice to hear about this,
but what about the post-gain issue of the ds-filter??
when will there be a new release??

I don't want to burn movies to cd and trust that there will be a working solution in the future!! First, a working filter, then I start burning!

Raist
BTW, greetings from an internet-café on Teneriffa:devil:

[Defiler]

Here's a 1.0 version of Oggenc:
http://www.saunalahti.fi/~cse/oggenc.zip

Saw that in this thread:
http://www.hydrogenaudio.org/forums/showthread.php?postid=23771

[DSPguru]

Vorbis v1.0 is now included in BeSweet v1.4b14 (http://besweet.notrace.dk) package.

Dg.

[Defiler]

Damn. Nice work, DSPGuru.
I JUST made a rip with the RC3 version of BeSweet last night. Just a few hours too early! Heh.

[DSPguru]

hehe :)

[kxy]

good work, will test :D

[kxy]

No one is excited as I am, and did some tests? I will leave the abx for later, right now I just did a file compasion test with 56 seconds of sample of drums beat and several high notes follow up.

07/11/2002 09:28 PM 999,613 DSPguru-499.ogg
07/11/2002 09:32 PM 999,708 headac3he-499.ogg
07/11/2002 09:20 PM 1,206,746 headac3he-old-rc3-499-lossless.ogg
07/11/2002 09:20 PM 970,892 headac3he-old-rc3-lossly.ogg
07/11/2002 09:34 PM 1,235,020 new-5.99.ogg
07/11/2002 09:33 PM 1,078,418 new-5.ogg
07/11/2002 09:34 PM 1,311,225 new-6.01.ogg
07/11/2002 09:34 PM 1,309,779 new-6.ogg

This is my personal option, no abx or anything. But I can clearly hear a big different when it goes to q6, it sounds more dynmaic! That is when the lossless coupling kicks in among with higher bitrates.

[unplugged]

Yes, I can confirm, -q6 seems the starting point for "vibrant" result with music, as happened and mentioned (www.hydrogenaudio.org) with past Vorbis beta and RC releases.

Actually I working to find the right edge around 5.50-6.00, since the bitrate variation isn't negligible.

5.6 seems very good (no ABX...)

P.S.: it's incredible that this last release (1.0) has even more (little) bitrate cut (with relation to 04/07 compile), almost noticeable at high -q values (>5), about ~1 Mb over 50-60 Mb CD sound.

[canadian_fbi]

hmm, i was under the impression that lossless coupling was sort of overkill compared with light lossy, like stereo vs. joint stereo in mp3...

but anyway, now lossless doesn't start until 6, so you're no longer held as strictly to 4.99 (the former sweet spot, imo). then i read on hydrogenaudio that the quality scale seems to have slid a bit, so that -q 4.5 now sounds like 4.0 in RC3 did. and now i read here that apparently lossless at 6.0 seems to sound a lot better (with a lot bigger filesize). my poor ears can't handle this :)

[Slogra]

Wow, this one sounds MUCH better than the last one i had on my hd. I'm not sure which version that was (i kinda overwrote it with the new one :rolleyes: ). The higher frequencies are much better now!!! :)

[PeterTheMaster]

i usually encode my movie audio at 0.05 which used to result in about 72kbps. now i tried the same audio again with besweet 1.4b14.
it resulted in 62kbps.

does this mean less quality or (far)better encoding algorithm?

[BlackSun]

RC4 was non-public and very good, that's why they rushed on 1.0 :)

[Jon Ingram]

Originally posted by PeterTheMaster
i usually encode my movie audio at 0.05 which used to result in about 72kbps. now i tried the same audio again with besweet 1.4b14.
it resulted in 62kbps.

does this mean less quality or (far)better encoding algorithm?
Well, you've encoded it. Listen to it - do you think the quality has improved, or worsened? The aim is to keep quality consistent across versions, although 'quality' is a very hard thing to define. That's why the 48kbits mode is quality -1, and not quality 0.

Personally, I think all of the low quality modes are *much* improved over RC3, so it's time to party (http://www.explodingdog.com/january1/brightcolors.html).

[canadian_fbi]

i don't trust my ears so much, so i ran a quick test just to determine the lowpass frequency cutoffs that vorbis 1.0 uses at various quality levels. here's my results:


-q bitrate cutoff
--- ------- ------
0 73.5 kbps 15.2 kHz
1 88.1 kbps 16.5 kHz
2 98.7 kbps 16.5 kHz
3 119.8 kbps 18.0 kHz
4 131.0 kbps 19.5 kHz (gradual)
5 168.7 kbps 20.0 kHz
6 199.6 kbps 20.0 kHz

this was done using cool edit pro for frequency analysis, oggenc for encoding, and oggdec for decoding, though i did the -q4 test using besweet just to make sure and i did get the same results. yes, this was a particularly difficult sample - i wanted to make sure it had plenty of frequencies up to 22 kHz. i personally can't hear past 18.5 kHz, so it seems that anything past -q 3 is overkill as far as frequencies are concerned. overall though i was pretty impressed at the amount of frequency information that was kept. almost too much so - i wish there were a way to cut out some of the stuff above 18.5 kHz at the higher q levels, since i can't really hear it anyway...

[Jon Ingram]

Originally posted by canadian_fbi
i wish there were a way to cut out some of the stuff above 18.5 kHz at the higher q levels, since i can't really hear it anyway...
Happily for you, there *is* a way. With the 1.0 encoder you can now tweak certain internal quality options -- lowpass being one of them.

To switch to 16.5kHz cutoff for example, append the following to the command line:
--advanced-encode-option lowpass_frequency=16.5
It's deliberately long, to stop people using it if they don't know what they're doing :)

Here (http://www.xiph.org/archives/vorbis/200207/0211.html) is the list of all internal options which have been exposed so far.

[canadian_fbi]

hmm, that is cool, but any chance of this being supported in besweet/oggmachine anytime soon? :)

and i did test the lowpass feature in headac3he just now (forgot about it) and it works ok - a cutoff of 14 khz for example results in a hard cutoff at 15 khz and -10 db between 14 and 15 khz. not sure if headac3he uses this feature in vorbis though or does a lowpass some other way.

[DarkAvenger]

HeadAC3he reads vorbis' low-pass setting for current config, so unless changed default low-pass values are used.
On encoding HAC3 tells vorbis to low-pass at the given setting. Everthing else is Vorbis' job. ;)

[Emp3r0r]

Does 1.0 have better multiple channel support?
Does Vorbis rival AAC when transcoding from 5.1 AC3?

[The People's Elbow]

Just encoded 2 audio streams @ Q0.01 ... awesome!!!
Finally we have a audio codec that produces acceptable audio quality at such a low bitrate and is able to be used in A/V compression. Very nice work dear developers, the difference to rc3 is amazing!

[kxy]

I use q6, I still don't understand why people who sacrafice that MUCH audio quality for file difference merely around 50 to 60 mb, that depends on the movie length and the material of course. But generally speaking, would this 50 to 60mb make that much visual difference if it is added to the videos encode?

edit: added the world visual

[The People's Elbow]

@kxy: I guess your post is a kind of response to mine, so I'll try to help you understand :) ...

1)With the newest release of ogg audio the sacrifice of audio quality has become much smaller IMO!

2)If you like movies with more than 1 audio track (me does!) you can multiplicate the space you safe at this low bitrate for acceptable video quality!

3)People who agree with 2) and aim to produce 1CD rips (me does! ;) ) will find out that the audio quality @64Kb now is really satisfying (At least IMO it is :) )

greetz, Elbow!

[PeterTheMaster]

elbow, we seem to do exactly the same thing.

[Alestrix]

elbow, we seem to do exactly the same thing.
Add me to the list :D

- A

[The People's Elbow]

:D I'm happy to hear that I'm not the only guy who's doing it this way! ;) Makes me feel safe somehow :P

greetz, Elbow!

[kxy]

Again, you people are not answer my question/challenge.

The question is:
Generally speaking, would this 50 to 60mb make that much visual difference if it is added to the videos encode?

The answer to that is not significant. But if it used in audio, it is like heaven to hell(q0 to q6). You will lose a lot of bit that is need for your audio. And from q0 to q6 you will be going from lossless stereo mode to a mix point stereo, as I don't think they use phase stereo anymore in vorbis 1.0. Now we all know lesser the bits lesser the audio quality. Since there is no argument there, let have a few words with stereo mapping.

Lossless uses Polar mapping, as it separates the stereo image into point and diffuse components and it decodes bit-identically to dual stereo, thus it saves the dolby downmix corretly and safely. In layman's words, it saves the surround information better.

Mix point stereo tends to collapse inward to a point somewhere within the stereo image. A practical example would be a gun shot in a large closed space. The Reverbeations and the ringing would centered towards the place where is gun is shooting at. To me, it will produce less resonance as opposed to the lossless mode. And with mix model, it is generally more aggressive in higher frequencies, thus lower the amplitudes to produce a 'nearly' in-phase sound.

So to me, you will not see much visual increase, but by increasing the audio, you will hear better thus have a better audio environment and in terms it makes you enjoy the movie better!

But then again, my argument will fail, if you just have the typical computer speakers, as you won't hear much of the difference; that goes the same with the headphone, after a long movie your ears will get *used to what your perceive, or after a loud explosion, you ear will hear less of the soft noise(example: speaking) right after.

[zoli]

@DSPguru:::

please put me in picture. if i downloaded the newest besweet that comes with libvorbis 1.0 and point oggmachine to that besweet.exe, then is oggmachine going to use 1.0 or still the rc3 dll-s? (or whatever it is technically correct to say)

[canadian_fbi]

Originally posted by zoli
please put me in picture. if i downloaded the newest besweet that comes with libvorbis 1.0 and point oggmachine to that besweet.exe, then is oggmachine going to use 1.0 or still the rc3 dll-s? (or whatever it is technically correct to say)

i'm not dspguru but since i've had my head in besweet the past couple hours i'll try and give you an answer...

if you unzipped the libvorbis.dll that comes in the latest release into your besweet directory (and overwrote any previous libvorbis that might be there), it should use 1.0. you can test this by playing the outputted vorbis file in winamp, and in the file info under "vendor" it should say something like "Xiph.Org libVorbis I 20020711".

[zoli]

thank you mr fbi.
indeed i was stupid enough to write before looking. i should have just checked out the files.

[The People's Elbow]

kxy wrote:
Again, you people are not answer my question/challenge.
- Sorry, but this doesn't make any sense and isn't true! If you had understood my post you would clearly detect that my answer to your question: "If it's worth the sacrifice of audio quality...! is YES! Though I described the circumstances (1CD Rip + Multilanguage) and it isn't a general answer for all rips.

Here just a stupid calculation:
90 Minutes Movie - 2 Audio tracks
---------------------------------
Audio tracks @ Q6 (I guess something about 192KBit/s) = about 250MB
Audio tracks @ Q0 (~64KBit) = about 85MB
If you use the second method the 1CD!! Movie gains 165MB of space! Instead of 550MB (in case of using xcd or bigger cd-r's) the video stream can consume up to 715MB... that's about 25% more space for video and you can bet that this difference is very obvious!

PS: The sound artifacts @Q0 are really not that bad - I even gave WMA@64KBit a shot earlier and it didn't ruin my rips at all. (though it doesn't sound nice and is no competitor to vorbis at its actual level)

greetz, Elbow!

[kxy]

Originally posted by The People's Elbow
Here just a stupid calculation:
90 Minutes Movie - 2 Audio tracks
---------------------------------
Audio tracks @ Q6 (I guess something about 192KBit/s) = about 250MB
Audio tracks @ Q0 (~64KBit) = about 85MB
If you use the second method the 1CD!! Movie gains 165MB of space! Instead of 550MB (in case of using xcd or bigger cd-r's) the video stream can consume up to 715MB... that's about 25% more space for video and you can bet that this difference is very obvious!

I usually don't post here unless I had done some testing. I had done about 10 movie audio tracks, the difference between q6 and q0 is around 50 to 80mb approx. That is all I am saying, not much visual quality if you only shave off ~65mb on average. And A LOT of difference in audio, at least to me.

But

Now if you want to include 2 audio track, then the difference will be 100 to 160mb, now you will seem a visual difference in encode. I will do the main one in q 4, other one(director's comment in q -1)


PS: The sound artifacts @Q0 are really not that bad - I even gave WMA@64KBit a shot earlier and it didn't ruin my rips at all. (though it doesn't sound nice and is no competitor to vorbis at its actual level)
greetz, Elbow!

Again, if you are saying WMA@64kbit is not bad. Then our hearing is very different, and our perception of good quality is different. Like I explained in detail from the the last post, in a typical computer speakers setup, you won't hear much difference anyway. If you dont' care about preserving the surround information and nearly identical downmix phase information, then no argument here.

[unplugged]

Originally posted by kxy
I use q6, I still don't understand why people who sacrafice that MUCH audio quality for file difference merely around 50 to 60 mb, that depends on the movie length and the material of course. But generally speaking, would this 50 to 60mb make that much visual difference if it is added to the videos encode?

1) It's a movie, without much music content Vorbis seems to "capitalize" better the bitrate (dialogs, background and FX at -q1/2 seem sounding very detailed).
2) Difference is often more than 60 Mb
3) Perception: audio lossy isn't audible as video lossy is, therefore some megs (60-90) could give a bit more "breath".
4) ...especially at Vorbis -q2 don't worry about quality ;), my setting.

[kxy]

Originally posted by unplugged


1) It's a movie, without much music content Vorbis seems to "capitalize" better the bitrate (dialogs, background and FX at -q1/2 seem sounding very detailed).
2) Difference is often more than 60 Mb
3) Perception: audio lossy isn't audible as video lossy is, therefore some megs (60-90) could give a bit more "breath".
4) ...especially at Vorbis -q2 don't worry about quality ;), my setting.

1) Movie soundtrack seems to me, that sometimes it can be more difficult than just music, depends on the material. For example, the movie soundtrack could be playing loud music in the background, a man speaking on the left, and then some other noise coming from the back. That is music + dialog + noise at different directions.
2) The difference would AVERAGE around 60mb(results taken from 10 movie sound tracks, could be more, could be less. 1.0 just came out, will do more testing and get better average...), but have you done any test on it, or just done some math calculation?
3) not just more breath, more dynamic, more impact, more resonace! Therefore give you a better environment for your movie enjoyment. Again, better surround phase info being kept and better surround downmix. See my example on the loud gun shot in a large hall.
4) Don't really understand what you are getting at here.

[The People's Elbow]

I guess kxy's point of view is very fixed, so it doesn't really make any sense trying to explain something to him...

PS:
kxy wrote:2) The difference would AVERAGE around 60mb(results taken from 10 movie sound tracks, could be more, could be less. 1.0 just came out, will do more testing and get better average...), but have you done any test on it, or just done some math calculation?
- There is not much to test... math is the only thing you need for bitrate calculations. Trying to archieve an average result for what??? Movies can be 80mins, 100mins, 200mins - the difference of q0/1/2 to q4/5/6/whatever files grows depending on the movie length - there is no sense in this action at all! Ogg uses quality levels and is vbr, but this q-levels are somehow equal to the mp3-abr mode though the output files tend to achieve a REAL bitrate that's a bit lower than the nominal bitrate (movie audio tracks at least seem to do that...)
... you don't have to encode an audio track to get to know the approx. difference between q-levels - that can be calculated.

cheers, Elbow!

[unplugged]

Originally posted by kxy
For example, the movie soundtrack could be playing loud music in the background, a man speaking on the left, and then some other noise coming from the back.

Right, for certain titles maybe mandatory use high -q values, (Gladiator, Blade Runner and some "vibrant" titles...) however I still point my impression that generic movie content doesn't fit solidly each second with sensible content as PURE music, almost unless you use very good listening system... (to be said most "sensible" titles are action or FX films).

Originally posted by kxy
That is music + dialog + noise at different directions.

Generally movie sound content is natural as smooth, and unless you normalize much, sound dynamics isn't so aggressive or demanding...
Again, in my opinion statistically the content isn't so rich-per-second or demading either for the encoder (codec, first of all tuned for music) or listener as a music-CD requires.

Originally posted by kxy
2) The difference would AVERAGE around 60mb(results taken from 10 movie sound tracks, could be more, could be less. 1.0 just came out, will do more testing and get better average...), but have you done any test on it, or just done some math calculation?
3) not just more breath, more dynamic, more impact, more resonace! Therefore give you a better environment for your movie enjoyment.

Pardon, I have said BIT more breath intending more to video bitrate.
Really I haven't tried -q6 for movie audio since old Vorbis compiles, so your 10 recent encodes can tell the fact :).

Originally posted by kxy
Again, better surround phase info being kept and better surround downmix. See my example on the loud gun shot in a large hall.

I don't use surround downmix, simply stereo with light normalization, but maybe if analog surround can be maintained after vorbis compression it's a good thing (Does -q6 manage to save the Analog Dolby rear channel mixed signal? :rolleyes: ).

[mustaneekeri]

Originally posted by raistlin2k
nice to hear about this,
but what about the post-gain issue of the ds-filter??
when will there be a new release??

Raist


what is the post-gain issue?

[kxy]

Originally posted by The People's Elbow
I guess kxy's point of view is very fixed, so it doesn't really make any sense trying to explain something to him...
cheers, Elbow!

Elbow, so far you haven't provide technical or hard fact to refute me. I like to get my conclusions from hard facts and listening results. To me, you are the person who would open up cooledit pro and analyze the graph and won't spend A lot of time on listening. Like I stated in the previous post, if you think wma 64kb is not bad, then we are in two different worlds. We have different standards. I am not saying your way is bad, if you are comfortable with that, then I am glad. :)

Again if you read my past posts, I said what I feel like, what I think, and how it sound to me therefore I am not forcing you to anything. I like to state my opinion.

Simply I want an answer on why people are not will to sacrifice a few bits for their video encode for a much better audio environment.

The file size of the audio you calculated is base on nominal rates, it can give an estimate but not exact target size, I would say around 10 percent.

[kxy]

Originally posted by unplugged


Right, for certain titles maybe mandatory use high -q values, (Gladiator, Blade Runner and some "vibrant" titles...) however I still point my impression that generic movie content doesn't fit solidly each [B]second with sensible content as PURE music, almost unless you use very good listening system... (to be said most "sensible" titles are action or FX films).


Yes, it is good that some agree with me. :D I am not saying all the movies should be q6. I use q6 as an extreme example for the FX films and as a black and white example as how it was compare to q0.

Generally movie sound content is natural as smooth, and unless you normalize much, sound dynamics isn't so aggressive or demanding...
Again, in my opinion statistically the content isn't so rich-per-second or demading either for the encoder (codec, first of all tuned for music) or listener as a music-CD requires.

I might be wrong here, so correct me if I am wrong. I thought that vorbis is very good at managing the bit rates. If the content isn't demanding, it naturally lowers it quite a bit. If it a poor source, as the background audio is hissing, then it might have pump up the bitrate a bit. That is why for most movies, the final target size is always lower than what I predict base on the calculation of the bit rates.

[QUOTE]
I don't use surround downmix, simply stereo with light normalization, but maybe if analog surround can be maintained after vorbis compression it's a good thing (Does -q6 manage to save the Analog Dolby rear channel mixed signal? :rolleyes: ).

Yes, another reason I used q6 as an example. When using q6, that is when the lossless coupling kicks in. It is very carefully implemented, thus it produce a nearly identically stereo image. Anything below q6 is lossly. Not saying lossly sucks, cause lossly coupling is really impressive too. Since version 1.0, the stereo modes are more complex in general from rc3. It just that in a few times that I can hear a warbling distorted sound from the rear speakers when I encode using lossly coupling. I can't exactly prove that coupling had the effect on surround. But from the few incidents that I have, doing a lossless coupling makes me feel much better. :O

I had asked monty(developer of vorbis for those who don't know) about this, here is the log, word for word....

<yaoyaoman> xiphmont: 2 questions. Let say my the A/V receiver is able to decode Dolby Surround - ProLogic II, now would there be a distinct difference if the vorbis is encoded with lossly channel coupling or lossless channel coupling? It was known that the old version of vorbi's bitrate allocation gives a long jump from -q 4.99 to -q5. Does the lossy channel coupling degradating the quality too much, and not becoming progressively less lossy and seamlessly lossless? In the RC3 versions, if I decoded my ac3 to lossy i can hear the warbling effect on the rear surround speakers if I force dolby sourround on my amp, it is less noticeble when it loseless.

<xiphmont> yaoyaoman: lossy stereo coupling will indeed affect it, but so will almost every other aspect of data compression int he codec. Also, the stereo changes are now closer to seamless. Not seamless yet, but closer. No longer a huge jump from 4.99->5. There's a jump yes, but more like at the other mode transitions.
<skip some other people's chat...>
<yaoyaoman> If one wants to save as much important surround(*hearable under a normal condition not with an audiophile equiment) information as possible then would you recommand lossly or lossless, given not too much file size increase.

<xiphmont> I have not tried it, so I do not know. 'least loss possible' will always give better results. If you want to take the time to test, we'd be interested in your results.

Back to your question, "Does -q6 manage to save the Analog Dolby rear channel mixed signal?" I can't tell you that. Since I don't have solid prove yet and I don't have a very solid understanding of how exactly coupling works and how it is scaled yet. But I can tell you base on my ear, q6 is damn good! So if the movie is very compressible and undersize, I would take an advantage of that as opposed doing a director's commentary.