Rediscovering Vorbis...

skottish · 2010-06-12 18:18:33

Some years back I encoded around 75 of my favorite CDs with Vorbis and was very disappointed with what I ended up with; There was this 'sonic hole' right in the middle of the spectral range that wasn't there with AAC-LC at similar bitrates. I ended up encoding my entire collection of then 950+ CDs in AAC-LC, bought an I-Pod, and didn't look back... until now.

I'm starting to look around for another audio player that isn't an Apple product and decided to try Vorbis again. Holy crap it sounds good! At -q 6 (~190Kbps), the sound is far beyond what AAC-LC can do and certainly is competetive with, if not superior to, AAC-HE encoded with neroAacEnc. I knew that there's been a number of significant improvements over the years with Vorbis, but I had no idea that it's become as excellent as it is. Very nice indeed!

Rasi · 2010-06-12 20:33:24

skottish wrote:

Some years back I encoded around 75 of my favorite CDs with Vorbis and was very disappointed with what I ended up with; There was this 'sonic hole' right in the middle of the spectral range that wasn't there with AAC-LC at similar bitrates. I ended up encoding my entire collection of then 950+ CDs in AAC-LC, bought an I-Pod, and didn't look back... until now.
I'm starting to look around for another audio player that isn't an Apple product and decided to try Vorbis again. Holy crap it sounds good! At -q 6 (~190Kbps), the sound is far beyond what AAC-LC can do and certainly is competetive with, if not superior to, AAC-HE encoded with neroAacEnc. I knew that there's been a number of significant improvements over the years with Vorbis, but I had no idea that it's become as excellent as it is. Very nice indeed!

And aotuv from aur is even better. (this one should not be used with anything lower than -q 6 tho - its not bad below, but its tweaked for +6)

skottish · 2010-06-12 21:04:32

From a quick search, it looks as if improvements made from aotuv are periodically merged into mainline. I've found more than one reference to code errors being discovered in aotuv whenever these merges have been or are about to happen. I'll have to do more research, but I'm wary of introducing anomalies into something as sensitive as audio compression.

It's cool that you brought up the -q 6 thing because I usually encode at the maximum quality for music. The last CD that I encoded to AAC averages something like 420Kbps. Overkill? Probably, but I have the space to handle it.

Runiq · 2010-06-12 21:48:36

skottish wrote:

Overkill? Probably, but I have the space to handle it.

That's why I encode to flac. I have an external 1.5TB drive where my music resides, and I use a script to copy it over to my portable audio player (calculates replay gain, mirrors directory structure, and re-encodes to ogg). That way I have the best audio quality and all the "space benefits" of a lossy format where I need them.

skottish · 2010-06-12 22:37:20

Runiq wrote:

skottish wrote:
Overkill? Probably, but I have the space to handle it.
That's why I encode to flac. I have an external 1.5TB drive where my music resides, and I use a script to copy it over to my portable audio player (calculates replay gain, mirrors directory structure, and re-encodes to ogg). That way I have the best audio quality and all the "space benefits" of a lossy format where I need them.

That sums up the direction that I've been moving. Re-encoding the CD library to FLAC will be a bit of a time consumer, but I simply didn't have the knowledge when I did it the first time to get it right.

I do like FLAC, and now Vorbis. Xiph has done a great job on their audio stuff.

Runiq · 2010-06-13 00:02:51

skottish wrote:

That sums up the direction that I've been moving. Re-encoding the CD library to FLAC will be a bit of a time consumer, but I simply didn't have the knowledge when I did it the first time to get it right.

Sorry, I'd post said script if I'd find it in this mess of data that is my back-up harddrive.

I do like FLAC, and now Vorbis. Xiph has done a great job on their audio stuff.

Yep, they sure have. When was it that you tried Vorbis first?

skottish · 2010-06-13 00:27:43

Runiq wrote:

skottish wrote:
That sums up the direction that I've been moving. Re-encoding the CD library to FLAC will be a bit of a time consumer, but I simply didn't have the knowledge when I did it the first time to get it right.
Sorry, I'd post said script if I'd find it in this mess of data that is my back-up harddrive.
I do like FLAC, and now Vorbis. Xiph has done a great job on their audio stuff.
Yep, they sure have. When was it that you tried Vorbis first?

I first used Vorbis when I started to use Linux full time. That was something like six or seven years ago. I've dabbled with it a bunch of times since then, but the last couple of days is when I really started to look at it seriously again.

stqn · 2010-06-13 01:16:23

I'd like to know at which quality you people start to hear compression artifacts... I've done blind tests on two albums (Emiliana Torrini's Me and Armini and Emily Haines and the soft skeletons' Knives don't have your back) and I could barely notice a minuscule problem at one point in one track of one of the two albums. At q-2.

So these days I'm usually encoding music at q2 but I think it's way overkill. I had a different experience when encoding the soundtrack of an anime, though; I had to go up to q1 to get decent results (for some sound effects and voices).

Last edited by stqn (2010-06-13 01:19:53)

fflarex · 2010-06-13 05:38:42

About a year ago (so... mainline libvorbis-1.2.x I think) I determined that -q0 was the point at which I started hearing artifacts in my music collection. Even then, it was only on a few problem songs. As a result I encode with -q1 and haven't had any issues with it.

My flac collection is only getting less and less relevant. I have contemplated deleting it, but there's no sense in that while I still have the disc space.

Runiq · 2010-06-13 06:57:45

stqn wrote:

I'd like to know at which quality you people start to hear compression artifacts... I've done blind tests on two albums (Emiliana Torrini's Me and Armini and Emily Haines and the soft skeletons' Knives don't have your back) and I could barely notice a minuscule problem at one point in one track of one of the two albums. At q-2.
So these days I'm usually encoding music at q2 but I think it's way overkill. I had a different experience when encoding the soundtrack of an anime, though; I had to go up to q1 to get decent results (for some sound effects and voices).

Hm… Doesn't Vorbis use high numbers for good quality? Did they change that some time ago?

Well it depends – If you have a decent soundcard, headphones/speakers and the music comes from a quality source, you'll probably notice the artifacts sooner. I haven't done any blind tests, I just used the default so far, and didn't hear any difference between -q 6 and -q 10. Using Audioslave – Heaven's Dead as a "benchmark" – love those crispy-clear drums.

stqn · 2010-06-13 17:16:00

Runiq wrote:

Hm… Doesn't Vorbis use high numbers for good quality?

Indeed.
q-2 is the lowest possible setting at approximately 32 kbps.
q-1 = 48 kbps
q0 = 64 kbps
q1 = 80 kbps
q2 = 96 kbps
q3 is the default (for oggenc), at 112 kbps.

As far as my equipment is concerned, I have decent headphones (Audio-technica A500) but am using my laptop internal sound chipset (Intel 82801DB-ICH4).

PS: my tests were made with the latest aoTuV at the time (a few months ago). The official xiph encoder can't go lower than q-1 at this time. Also I think ffmpeg's vorbis encoder was/is very crappy so you shouldn't use this... (I think it's what the video encoder SUPER uses under Windows, and the results were extremely bad when I tried. Also read something about that recently on monty's blog...)

Last edited by stqn (2010-06-13 17:20:54)

Army · 2010-06-13 18:20:55

stqn wrote:

I've done blind tests on two albums (Emiliana Torrini's Me and Armini and Emily Haines and the soft skeletons' Knives don't have your back) and I could barely notice a minuscule problem at one point in one track of one of the two albums. At q-2.

Honestly?!?!
Well, I can hear differences very easily at -q4, -q5 is really much harder and -q6 is transparent for me. So I don't recommend going > 6, it's just a waste of disc space, before you go up to -q10 you really should use flac!

I've been using Aotuv's version for the past few years, even in my Windows XP times. Right from the beginning I had a vorbis capable player (iAudio G3 and afterwards iPod Video with Rockbox, now Sandisk Sansa e260 with Rockbox => Rockbox is awesome, it now even plays atrac3!! *lol*). Now I have converted my whole collection to flac and it's really cool to have it in the original quality, the only real way to backup your CDs!
Now I switched to the official version of vorbis, because there was an update in Arch which made Aotuv's version fail to build. But his patches go into the official version, so you really have a very good quality with libvorbis and vorbis-tools from the repos

skottish · 2010-06-13 18:45:44

Even at -q 10, you'll still end up with a file size much smaller than what flac produces. It is overkill though. I need encoded files for my portable, as well as for videos that I make. Happily, DivX supports vorbis, so it makes it easy for me to get my artwork out to my friends using primitive operating systems.

Larsson · 2010-06-14 18:32:16

For making a backup of a cd, the best way is to rip it with EAC (via wine) to FLAC, and there are some guides on the internet that makes it easy to set up. You get a log and cue so it's easy to burn a perfect copy of your cds.

If one only are interested Vorbis, then rubyripper is a very good alternative, and then vorbis att -q6.

.:B:. · 2010-06-14 20:27:49

Rasi wrote:

And aotuv from aur is even better. (this one should not be used with anything lower than -q 6 tho - its not bad below, but its tweaked for +6)

That's funny... My understanding has always been aoTuv was actually optimised for lower bitrates, not higher. Why optimise for higher bitrates? The higher your bitrate the less information you'll cut out anyway. Makes no sense.

If you can back it up I'm happy to adjust my opinion of course .

Labello · 2010-06-14 20:28:22

Are you guys really serious about the fact that you hear differences between the different bitrates? no offence i am really curious! maybe my sound-system i crap, but i don't hear any differences between different bitrates... well between 32kbit and 320kbit i do but not within such small margins o__O

Zeist · 2010-06-14 20:35:27

I do like vorbis for portable players like my N800 or my iAudio X5L since then it's transparent. On my HTPC connected to my home theatre system however even vorbis -q10 isn't good enough to be entirely transparent to lossless so in that case I stick with Wavpack.

Larsson · 2010-06-14 20:35:28

Labello wrote:

Are you guys really serious about the fact that you hear differences between the different bitrates? no offence i am really curious! maybe my sound-system i crap, but i don't hear any differences between different bitrates... well between 32kbit and 320kbit i do but not within such small margins o__O

No, but it makes me feel very fuzzy inside, knowing that my music is prefectly stored in FLAC

Last edited by Larsson (2010-06-14 20:35:44)

Runiq · 2010-06-14 20:51:35

Larsson wrote:

Labello wrote:
Are you guys really serious about the fact that you hear differences between the different bitrates? no offence i am really curious! maybe my sound-system i crap, but i don't hear any differences between different bitrates... well between 32kbit and 320kbit i do but not within such small margins o__O
No, but it makes me feel very fuzzy inside, knowing that my music is prefectly stored in FLAC

Yeah, same here. I sleep rather well at night.

karoshi · 2010-06-14 20:53:58

Labello wrote:

Are you guys really serious about the fact that you hear differences between the different bitrates? no offence i am really curious! maybe my sound-system i crap, but i don't hear any differences between different bitrates... well between 32kbit and 320kbit i do but not within such small margins o__O

Well, the experience of music is not always limited to what you hear with your ears. some music you can feel with your skin or in your stomach (deep, heavy basslines for example). mp3s, even encoded with 320kbps, cant give you that experience, imho. the punch gets lost on compression. (happy ogg -q10 user here)

Rasi · 2010-06-14 21:11:38

.:B:. wrote:

Rasi wrote:
And aotuv from aur is even better. (this one should not be used with anything lower than -q 6 tho - its not bad below, but its tweaked for +6)
That's funny... My understanding has always been aoTuv was actually optimised for lower bitrates, not higher. Why optimise for higher bitrates? The higher your bitrate the less information you'll cut out anyway. Makes no sense.
If you can back it up I'm happy to adjust my opinion of course .

http://wiki.xiph.org/VorbisEncoders#AoTuV

"Vorbis: Vorbis is now impressive. Last year constant noise boost or coarseness
spoiled the performance of this format. I was surprised to hear those problems
on –q6,00 setting supposed to be free of them due to lossless coupling. Aoyumi has
apparently identifies precisely the cause of this problem and he worked to solve it.

Last edited by Rasi (2010-06-14 21:13:00)

.:B:. · 2010-06-15 05:46:28

Bah!

Then wikipedia must be utterly wrong:

The Xiph.Org Foundation maintains a reference implementation, libvorbis, the latest official version of which is 1.3.1, released on March 26, 2010[4]. There are also some fine-tuned forks, most notably aoTuV, that offer better audio quality, particularly at low bitrates. These improvements are periodically merged back into the reference codebase.

And a second mention of it being focused on lower bitrates.

I don't see how your quote proves aoTuv is tuned for higher bitrates. Note it says 'higher quality', not 'higher bitrates' :

aoTuV encoders are optimized encoders for higher quality.
* aoTuV beta2 was adopted as a part of Vorbis 1.1 of Xiph.Org (quality wise all versions of official libraries from 1.1 up to 1.2.1 are the same of aoTuV beta2).
* aoTuV beta4 improved quality at all bitrates (versus xiph.org 1.1.2) and can also go down to 32 Kb/s (q-2). Note that this patch also include the Vorbis-OptSort patch.
* The latest release is aoTuV beta5.7.
* The latest stable release is Release 1 (AKA aoTuV beta 4.51), which improves quality at q < 3 (versus aoTuV beta4).

The llines you quote talk about artifacts that the author did not expect to hear at Q6 anymore but were still there (and subsequently fixed). That's however no indication that it's tuned for Q6 if you ask me.

I think we'll have to meet in the middle .

fflarex · 2010-06-15 06:13:22

I'm pretty sure that aoTuV is tuned for classical music at low bitrates. However, I don't have a source as I heard this so long ago and it was probably just from a random forum post at hydrogenaudio. It's highly possible that the emphasis on classical music was dropped at some point or that I just heard wrong. The quote that Rasi posted sounds to me like it's just about a bugfix, though. It makes no sense to optimize for high bitrates. Any encoder under the sun can achieve high quality with high bitrates!

Army · 2010-06-15 07:37:17

To get this straight, Aotuv is optimized for lower bitrates. But all in all it has a better quality on every bitrate, so you can say it's more efficient. Also, in Aotuv the bitrate of e.g. -q4 is closer to 128kbps than in the official version, which was always far below. So all in all Aotuv gives you a much better quality, when you use e.g. -q4, because 1. you get a higher bitrate and 2. Aotuv is better at the same bitrate, so its quality is even much better because of 1.

But these times are gone, I don't hear or see (bitrate comparison etc) any real differences anymore between the official version and Aotuv, so right now I stick with the version from the repos. And even if there will be a huge quality improvement in the future, I don't worry, because I have everything encoded in flac :-) (I only encode stuff to vorbis when I put it on my rockbox player)

Last edited by Army (2010-06-15 07:52:27)

dyscoria · 2010-06-15 11:17:09

What i love about encoding in FLAC is that the rips can stand the test of time.... can convert FLAC back to WAV without any loss of quality and then convert the WAV to any format you wish. If you have an OGG or MP3 rip, you are stuck with that format.

Arch Linux

#1 2010-06-12 18:18:33

Rediscovering Vorbis...

#2 2010-06-12 20:33:24

Re: Rediscovering Vorbis...

#3 2010-06-12 21:04:32

Re: Rediscovering Vorbis...

#4 2010-06-12 21:48:36

Re: Rediscovering Vorbis...

#5 2010-06-12 22:37:20

Re: Rediscovering Vorbis...

#6 2010-06-13 00:02:51

Re: Rediscovering Vorbis...

#7 2010-06-13 00:27:43

Re: Rediscovering Vorbis...

#8 2010-06-13 01:16:23

Re: Rediscovering Vorbis...

#9 2010-06-13 05:38:42

Re: Rediscovering Vorbis...

#10 2010-06-13 06:57:45

Re: Rediscovering Vorbis...

#11 2010-06-13 17:16:00

Re: Rediscovering Vorbis...

#12 2010-06-13 18:20:55

Re: Rediscovering Vorbis...

#13 2010-06-13 18:45:44

Re: Rediscovering Vorbis...

#14 2010-06-14 18:32:16

Re: Rediscovering Vorbis...

#15 2010-06-14 20:27:49

Re: Rediscovering Vorbis...

#16 2010-06-14 20:28:22

Re: Rediscovering Vorbis...

#17 2010-06-14 20:35:27

Re: Rediscovering Vorbis...

#18 2010-06-14 20:35:28

Re: Rediscovering Vorbis...

#19 2010-06-14 20:51:35

Re: Rediscovering Vorbis...

#20 2010-06-14 20:53:58

Re: Rediscovering Vorbis...

#21 2010-06-14 21:11:38

Re: Rediscovering Vorbis...

#22 2010-06-15 05:46:28

Re: Rediscovering Vorbis...

#23 2010-06-15 06:13:22

Re: Rediscovering Vorbis...

#24 2010-06-15 07:37:17

Re: Rediscovering Vorbis...

#25 2010-06-15 11:17:09

Re: Rediscovering Vorbis...

Board footer