Which program produces better WAVs : Azid or Vob2Audio

[iparout]

Hi there.

I have a question for the more experienced users here...

I have been using VBR mp3 as the audio for my DivX encodings from DVD Ripps and the software I use to encode to mp3 is Lame. The source file for the mp3 is a WAV file created by Vob2Audio.

What I would like to know is this : If I use DVD2AVI to demux the ac3 file and then Azid to transcode to WAV, will the WAV file be better in quality than the method I use ?

Thanks in advance.

[DJ Bobo]

AZID is rumored to be the best surround downmixer available. So I guess yes.

[pacohaas]

Originally posted by iparout
If I use DVD2AVI to demux the ac3 file and then Azid to transcode to WAV, will the WAV file be better in quality than the method I use ?You'll get a faster conversion and better quality(because of a completely floating point process) by using BeSweet to go directly from the vobs or ac3 to mp3. It uses azid and lame all in 1 go, so it's exactly what you are looking for.

[DJ Bobo]

Well, even you don't use BeSweet, you can still add

-F wav24

to your azid command, which will give you maximal quality.
And the one that is going to demonstrate that a 100% floating point processing sounds better isn't born yet!

[iparout]

So which is the best solution in terms of quality.

Use BeSweet to directly transcode AC3 to mp3 or
Use Azid for AC3 --> WAV and then Lame for WAV ---> mp3 ?

Or is it the same ?

Keep in mind that I also normalize the WAV file using Sonic Sound Forge. I know that the tools you suggest also support on-fly normalization but I don't know how good this method is implemented by those tools.

@forum:Which method do you guys use ?

Thanks.

[DJ Bobo]

Better do the normalisation within AZID using the -a switch (it is done in floating point, so best quality anyway)
A typical azid command looks like this for me:

5.1 sound:
azid -c normal -L -3db -a -F wav24 name.ac3 name.wav

2.0 stereo or surround sound:
azid -c normal -a -F wav24 name.ac3 name.wav

2.0 mono or 1.0 sound:
azid -c normal -a -F wav24 -d1/0 -oc name.ac3 name.wav

And then encode with Lame.
The quality will be the same as with besweet (I just don't use BeSweet because I feel better doing things manually and controlling each point of the conversion)

[iparout]

First off, thank you for helping me out.

Now, I have downloaded BeSweet GUI and the default azid command is :

-azid -n1 -c normal -g 0db -L -3db

So I have some questions :

1) What is Dialogue Normalization in Azid ? (-n1 switch.)Is it helpful ?

2) If I want to Normalize to 0 db (100%), which switch should I add ? I think it's the --maximize switch which is the same as the -a switch you told me, but I need confirmation on that. Any difference between those two switches ? (they appear as the same switch in the Azid .txt)

3) I can see that you put the downmixing (-L -3db) before the normalization (-a). Is there a special reason for this or is it just random ?

4) Can someone suggest a good Lame switch. I have been using --alt-preset 128 but Besweet suggests also using --scale 1 -p as well, which stands for Error Protection and Scale Input. Can someone explain what those two commands do and if they are useful ?

5) Last but not least, does BeSweep create a temp .WAV file when transcoding from ac3 --> mp3 ? If no, then why bother include the -F wav24 switch ( = generate 24-bit floating point WAV) if no WAV is actually generated ?

Thanks in advance for any help.

[DJ Bobo]

1) I don't know about the -n1 switch and I never use it. I can't help you here.

2) --maximize = -a (exactly the same thing)

3) you can put the -L -3db after -a. I just feel better putting it before :D

4) error protection isn't needed, it is mandatory, do as you like here (I never bother using it). As for a good lame setting, I'm a great fan of the "--r3mix" preset. It produces perfect quality audio and nevertheless keeps the bitrate somewhere between 128 & 160kbps.
--r3mix is equivalent to: --nspsytune --vbr-mtrh -V1 -mj -h -b96 --lowpass 19.5 --athtype 3 --ns-sfb21 2 -Z --scale 0.98 -X0
I personally tweak it that way:
* if I have a 448kbps 5.1 AC3 or 2.0 streams, I change the lowpass to: --lowpass 20.3
* if I have a 384kbps 5.1 AC3, I change the lowpass to: --lowpass 18.1
* You can save 10kbps on the average by changing the V and b settings to: -V2 and -b32. This won't cause any hearable quality loss (unless you have a very high end system and so called golden ears! ;))
As for mono streams, I just use -b 32 -m m -h -V 0 -B 160
This will give you 96 to 128kbps average and perfect quality! you can of course go a little bit down with the V setting, down to 1 or 2 for example.

5) Don't know what BeSweet does or not. I think DSPguru will be answering this :D

[pacohaas]

for recommended LAME settings please read the FAQ, r3mix hasn't been recommended for quite some time now...better settings are definately out there and they have been proven time and again through blind listening tests to be better. If you want the filesize of r3mix and higher quality, use the 2-pass method I explain here (http://forum.doom9.org/showthread.php?s=&threadid=33039).

@DJbobo, maybe you mistyped, but "mandatory"=you must use it. I think you mean "optional".

[DSPguru]

Originally posted by DJ Bobo
Don't know what BeSweet does or not. I think DSPguru will be answering this :D i have golden ears, so i guess i won't even bother.

[DJ Bobo]

@ pacohaas
I already know about --alt-preset and use it only if I want to reach a given average (low) bitrate. I also know about the --alt-preset xtreme, insane & co settings, but they produce higher bitrates than r3mix and I don't hear any difference, that's why I don't use them.

The 2-pass method takes too long time. Besides, I don't believe --alt-preset xyz will give better quality than r3mix. This is not logical, r3mix is true VBR where --alt-preset will be encoding not far away from the average bitrate and won't act dynamically enough to soft or hard audio pieces.

And no I havn't mistyped, I really thought that "mandatory" = "optional". Sorry! :D

@ DSPguru
You misunderstood me may be. Point5 was an answer to his question #5. He asks about the functioning way of BeSweet, so I thought you were more suited to answer his question.
No offence intended.

[pacohaas]

Originally posted by DJ Bobo
Besides, I don't believe --alt-preset xyz will give better quality than r3mix. This is not logical, r3mix is true VBR where --alt-preset will be encoding not far away from the average bitrate and won't act dynamically enough to soft or hard audio pieces.I know it seems illogical, i totally agree and was against moving away from my custom VBR line to an abr line for low bitrates, and perhaps at the bitrates you are looking(128-160), VBR could still be better than ABR, but i know many people over at hydrogen would strongly discourage you from using the very outdated --r3mix.

If it works for you, fine, but don't recommend it to others as a good high-quality setting, when you know there are better ones out there.

[DJ Bobo]

As I always say, there is "better quality" and "better quality". There is better quality that is hearable, and there is better quality that isn't hearable. Only the first one is relevant for me. So following this, I assume --r3mix *IS* a HQ setting, as it delivers excellent results and the --alt-preset VBR presets aren't any better in my ears (and I'm not using crappy PC speakers!). Add to this that r3mix achieves lower bitrates ;)

I tell you when I use --alt-preset:
1) When I need a low bitrate between 80 and 128kbps, I use --alt-preset xyz
2) When I rip Audio-CDs and need compromise less, state of the art quality, I use --alt-preset insane

For all other things, I use --r3mix or its tweaked versions as above.

The last option remaining is CBR encoding, which I use if I don't have time or if I just don't care for a particular title.

All I wanna say is: I'm realistic and don't throw people in panic by describing everything else than the state of the art settings as crap. I just see things as they are and I'm sure you'll agree with me that what I'm saying is very near to the reality and has nothing to do with pure theories.

[iparout]

So, DSP Guru, could you please answer to my 5th question first ? Seems like you are the most appropriate guys for it.

[pacohaas]

Originally posted by DJ Bobo
As I always say, there is "better quality" and "better quality". There is better quality that is hearable, and there is better quality that isn't hearable. Only the first one is relevant for me......The last option remaining is CBR encoding, which I use if I don't have time or if I just don't care for a particular title.

All I wanna say is: I'm realistic and don't throw people in panic by describing everything else than the state of the art settings as crap. I just see things as they are and I'm sure you'll agree with me that what I'm saying is very near to the reality and has nothing to do with pure theories. fair enough, as I eluded to before, i used to use r3mix for the longest time when roel was actively tweaking it with new versions of LAME. I'm not saying r3mix is crap, just that things have evolved and it has not. It's still far better than using audiocatalyst vbr or wma or CBR(there's always time to use ABR), so i think we're pretty much on the same side here ;) If you could just mention the newer presets as the alternative that most people use, i think i'd agree more with your posts.

[DSPguru]

Originally posted by iparout
does BeSweep create a temp .WAV file when transcoding from ac3 --> mp3 ? If no, then why bother include the -F wav24 switch ( = generate 24-bit floating point WAV) if no WAV is actually generated ?BeSweet will never create a temp wave file and will always work on float-mode (as indicated in the logfile).
so that makes "-F wav24" redundant.

btw, iparout, you might also wanna search the forum for BeSweet's unique normalization mode called one-pass normalization.

[iparout]

Originally posted by DSPguru
btw, iparout, you might also wanna search the forum for BeSweet's unique normalization mode called one-pass normalization.

I searched the forum for what you suggested but couldn't find any answers. Anyways, I use Autofind Maximum Gain in Azid so I guess that the option you are suggesting wouldn't do anything better for me...

However, I still got that question about Dialogue Normalization and Scale Input you have enabled in your profiles... I am sure as hell I don't hear any difference inthe output file... (But of course, I have $5 speakers...:D) Can you please tell me what they are *supposed* to do and whether they result in better quality ?

Thanks in advance.

[frank]

Dialogue level is an important value in professional audio editing!
It is the reference, the centre channel is fixed to -31dB. Professionals don't work with normalization BS.

The dialogue level parameter in AC-3 streams indicates the ENCODED level, mostly -27dB on DVD tracks.
If you set dialogue normalization = ON then AZID lowers the sound level of all channels until the centre channel (dialogue channel) has -31dB average loudness.

So every program content gets the same loudness (-31dB), independently from the ENCODED level. And there is enough headroom (= 31dB) for peakes, gun fire, explosions...

Example: If you change the DVD in your player you'll ever get the same loudness on AC-3.