libsamplerate/libvorbis: newer versions are available

[Dukers]

libsamplerate
Trunk code uses libsamplerate 0.1.4 (link).

The current upstream version is 0.1.7:

http://www.mega-nerd.com/SRC/download.html
http://www.mega-nerd.com/SRC/ChangeLog


libvorbis
Trunk code uses libvorbis aoTuV beta5 (link).

The current upstream version is beta5.7:

http://www.geocities.jp/aoyoume/aotuv/

Thank you.

[JohnAStebbins]

The development forum isn't the proper place to make requests. But is there some specific problem or feature enhancement that either of these libraries address? Without some justification, there's not a lot to motivate anyone to update them.

[Dukers]

Sorry if I posted at wrong place.

Well, the new libvorbis has some tunings at low bitrate and a security bug fixed.

BUT, this already have been posted here:

http://forum.handbrake.fr/viewtopic.php?f=6&t=9792

Next time better use forum's search function

[Dukers]

Now version 0.9.4 was released, maybe someone can take a look at these libraries for next release. And I add LAME to the list (and maybe libogg?).

I think it's important have AAC alternatives updated, because for Windows and Linux users (us poor mortals without Core Audio), FAAC's quality notoriously isn't so good.

[rhester]

But you think LAME can produce superior results to FAAC at sane bitrates?

Rodney

[JohnAStebbins]

We did a blind listening test on #handbrake a couple months ago. Dark_Shikari prepared samples using same bitrate of nero aac, vorbis, lame. He also supplied a lossless flac for comparison. For fun, I guessed which was which then rated them in quality. To my ear, highest to lowest was flac, vorbis, aac, mp3. Unfortunately, we didn't throw faac into the mix.
http://handbrake.fr/irclogs/handbrake/h ... 1_pg2.html

[rhester]

I would expect FAAC to fall between Nero AAC and MP3, depending on bitrate. Anything above 160kbps ABR is unachievable with FAAC (obviously).

CoreAudio AAC I'd expect to be quite competitive with Nero AAC, largely owing to the latter's lack of updates in over a year.

Rodney

[Deleted User 11865]

I would expect FAAC to fall between Nero AAC and MP3, depending on bitrate. Anything above 160kbps ABR is unachievable with FAAC (obviously).

Doesn't HandBrake use CBR for all encoders except Core Audio?

[Dukers]

FAAC is pure crap. It's dead. No improvements in quality for years (or maybe decades? ). Unfortunately it's the only free (as free speech) AAC encoder available.

And yes, LAME quality is better than FAAC even at 128kbps if VBR is used (Handbrake uses CBR). Vorbis is superior, close to a good AAC encoder.

BTW, there's a new version of Nero AAC encoder:
http://www.hydrogenaudio.org/forums/ind ... opic=77195

[Dukers]

http://trac.handbrake.fr/browser/trunk/ ... lame.c#L51

Changing line 51 to something like that (I'm not a programmer)

lame_set_VBR(gfp, vbr_abr);
lame_set_VBR_mean_bitrate_kbps(gfp, (<bitrate>));

produces an ABR file, doesn't it?

Took from here: http://lame.cvs.sourceforge.net/viewvc/ ... iew=markup

[TimmyC]

I realize this post is a month-later resurrect of this thread, but I figured that was better than starting a new thread.
I would love to see the LAME mp3 encoder implemented in its full-quality (VBR presents) form in Handbrake.
In blind testing done by Hydrogen Audio, Lame's current implementations have performed quite well for quality audio encoding, with quality usually not statistically differentiable from aac or ogg vorbis.

That presumes two things:
1) high-quality aac implementation, whether from Apple or Nero
2) Lame's VBR preset encoding, for which the Lame encoder has been optimized over the years

Given the oft-mentioned age and "unimpressive performance" of faac, I suspect pretty strongly that Lame's VBR implementation (at comparable bitrates) would outperform FAAC.

According to this thread, avi has trouble dealing with VBR which is the reason for enforced CBR audio encoding in Handbrake.
http://forum.handbrake.fr/viewtopic.php ... 80&p=30414

But especially now with 0.9.4, with Handbrake moving away from .avi compatibility, it seems like this shouldn't be a big deal.
Also, Handbrake has moved toward "VBR" video encoding (quality target instead of bitrate target) so it is already embracing some of the reasons behind VBR.

Of course, it's possible that VBR audio streams are still problematic for video container formats, in which case my point is moot.
But if VBR audio streams aren't intrinsically problematic, then implementing settings for Lame VBR audio streams would be a good thing, if Handbrake is going to have an option for mp3 audio stream in the first place.

[Dukers]

TimmyC, you're absolutely right. Unfortunately I don't have the skills to play with this.