Log in

View Full Version : Converting AC3 to vorbis - maximal quality


bkman
17th June 2006, 13:09
Can anyone advise me as to how I might convert AC3 to ogg vorbis while maintaining the maximum range and quality of sound? I've tried foobar with the ac3 component, but I can only use it in 16-bit mode as it seems to have a bug where it won't output 32-bit audio after 1:33 minutes.

Awatef
19th June 2006, 15:25
I use HeadAC3he. It supports floating-point processing.
Homepage of the program: http://mitglied.lycos.de/darkav/

Note that the SSE2 DLLs also work on Celeron-M, Pentium-M (Centrino), Celeron D and Athlon64 processors (and some others I don't recall here).
So use these if you have an adequate CPU for better processing speed.

bkman
19th June 2006, 16:29
Thanks. That seems to do what I need.

SeeMoreDigital
19th June 2006, 17:17
Kurtnoise13 is constantly developing his BeLight application...

http://kurtnoise.free.fr/BeLight/BeLight.gif

It provides direct AC3 to Vorbis encoding and offers many tweaks ;)


Here's a link: http://forum.doom9.org/showthread.php?t=85566


Cheers

unskinnyboy
19th June 2006, 17:49
I'd go with BeLight/BeSweet too. More options than HeadAC3he especially since HeadAC3he isn't actively developed anymore. Last installment that I know of HeadAC3he is from May '04. Get it here (http://forum.gleitz.info/attachment.php?attachmentid=66899&d=1085577165). This version is not available for download from DA's website.

BeSweet also uses floating-point arithmetics for transcoding. HeadAC3he is known to be a bit faster at the expense of needing more harddisk space for the temp WAV files it creates. Here (http://forum.doom9.org/showthread.php?t=12092) is a comparison thread.

gameplaya15143
19th June 2006, 18:19
I'm not sure, but don't all existing vorbis encoders only support 16bit input? I think that is the question that is being asked here.

16bit provides ~96dB of dynamic range. To actually use that much range, would give you hearing damage. Just use belight or headac3he with low or no DRC. Also, use the latest aoTuV when encoding in vorbis. (you can get one for headac3he on rarewares.org)

SeeMoreDigital
19th June 2006, 19:12
I'm not sure, but don't all existing vorbis encoders only support 16bit input? I think that is the question that is being asked here.

16bit provides ~96dB of dynamic range. To actually use that much range, would give you hearing damage.Hmmm!

What would be the point of altering the audio sample size anyway?

bkman
19th June 2006, 19:43
Belight doesn't seem to allow 32-bit output, and I've just found that BeSweetGUI does (I'm not well versed in the CLI settings). It gives me downmix overflows, however, and from what I have read HeadAC3 has better normalization.

Maybe it's just my imagination, but I can the difference between 16 and 32-bit processed versions of my audio source.

SeeMoreDigital
19th June 2006, 19:48
Maybe it's just my imagination, but I can the difference between 16 and 32-bit processed versions of my audio source.Once again...

What would be the point of altering the audio sample size anyway?


Cheers

Kurtnoise
19th June 2006, 19:52
Belight doesn't seem to allow 32-bit output
Wrong...BeLight can do that.

however, and from what I have read HeadAC3 has better normalization.
Give me some proofs please.

Maybe it's just my imagination, but I can the difference between 16 and 32-bit processed versions of my audio source.
I guess you mean "I can hear difference" ? If yes...prove it.

bkman
19th June 2006, 20:05
Wrong...BeLight can do that.

Can you show me where I can enable it? I don't see how to even output 32-bit STEREO WAV.

Give me some proofs please.
I guess you mean "I can hear the difference" ? If yes...prove it.

Yes, I did mean that, but it is not for me to provide proofs because I am not claiming anything. I am simply telling my current beliefs and conclusions. If they are wrong, then you can disprove them and educate me (if you want).

Kurtnoise
20th June 2006, 08:54
Can you show me where I can enable it? I don't see how to even output 32-bit STEREO WAV.
oOo weird...You didn't mentioned stereo output in your last reply. So, how can I guess that ? But here is a trick : select 32 bits 6 mono wavs and merge L/R channels with SoX.


Yes, I did mean that, but it is not for me to provide proofs because I am not claiming anything. I am simply telling my current beliefs and conclusions.
You didn't claim but YOU wrote those sentences. So, YOU must prove something to other people around here, not me. In this way I'm little bit curious to know how you perceive those things. That's why I'm asking your method to etablish this...

Anyway, many thanks for this useless post.

vlada
20th June 2006, 11:38
bkman> If you think you can hear something, try to ABX it. You can find more information at HydrogenAudio.org (http://hydrogenaudio.org).

I'm not sure, what version of libvorbis.dll is included with BeLight. For maximum quality you should use AoTuVb4.51. Get it here (http://rarewares.org/ogg.html).

bkman
20th June 2006, 11:45
oOo weird...You didn't mentioned stereo output in your last reply. So, how can I guess that ? But here is a trick : select 32 bits 6 mono wavs and merge L/R channels with SoX.

That seems like unneeded work when HeadAC3 does it in one step. Besides, I don't see Belight set the azid "-F wav32" flag, so I assume that it would be working with 16-bit decoded AC3

You didn't claim but YOU wrote those sentences. So, YOU must prove something to other people around here, not me.

That is an odd (and silly) way of looking at it. Do you expect everyone who posts to prove everything they write? How about you prove that I need to prove what I wrote? My belief that I can hear the difference is not what this topic is about. I was simply asking for a tool that can re-encode AC3 with maximum precision.

In this way I'm little bit curious to know how you perceive those things. That's why I'm asking your method to etablish this...

16-bit processed AC3 simply sounds less "vibrant" than the fully 32-bit equivalent for my audio (to me). I'm not trying to convince anyone else that this is possible, as it may just be a placebo effect. Either way, I would be happier with higher-precision processing.