Audio Encoding Options?

[whcanilang]

I'm using AutoGK 1.80 beta, updated with Lame 3.96.1 and VDubMod 1.5.10.1 build 2439.

Is it possible to alter the BeSweet options so that the Audio is encoded to Joint-Stereo ["-m j"] (rather than default Dual Channel) and VBR 130 [actually I'd like "-v --vbr-new -V 5" since VBR sounds better than ABR and vbr-new is faster than vbr-old] (rather than default ABR 128)???

I've already swapped my lame-enc.dll to use 3.96.1, tried editting the Defualt and Lame Default profiles in BeSweet's DD-Profiles.ini, and inseted an AutoGK profile into same. The profiles were all ignored so I assume AutoGK keeps it profile elsewhere.

I tried recoding the audio while VDubMod, but the final file was too large for a CD. I think it was large by the small diff in sizes btwn the ABR 128 file vs. the VBR 130.

Is this possible, or do I need to wait for a user-config file in the future??? Thanks and great bundle!!!

[jggimi]

You can do this sort of custom work right know with GK, if you wish. But to do this in AGK would obviously require development.

Moved to development forum.

[whcanilang]

Thanks. I thought it was possible in GK, but I _really_ appreciate the one-click convenience on AutoGK.

Any idea on where this might be on the AutoGK development priority list and approximately when it might be addressed?

[len0x]

Originally posted by whcanilang

Any idea on where this might be on the AutoGK development priority list and approximately when it might be addressed?

It looks like a feature for config file. It might be a milestone of its own - but timing is defined by how quickly current milestone target will be reached (so that new one can be defined). Read about milestones on the AutoGK's web site.

[Gambit]

I have to bump this thread a bit.

This is IMO more like a bug that needs to be fixed. Using dual channel is simply a bad idea. Plus, I've noticed that the AutoGKnot produced mp3 file has a different duration. Seems there is something bork in the encoding process.

Also, the used terminology is simply wrong, since VBR in AutoGknot is in fact ABR. I would suggest that for the settings you could choose ABR (and a target bitrate) or VBR (and the presets via the -V0-9 switch ).

[whcanilang]

Yes, from all I've read, Joint Stereo is a better use of bits than Dual Channel. Also VBR is supposed to be superior to ABR. I've also read that ABR is a sub-type of VBR, so this is a bit of a terminology confusion.

However, this does not rise to the level of a BUG! These are merely design choices that I wish were not hard-coded in, but configurable either through a GUI or a config file.

Thanks!

[Gambit]

I didn't mean to be offensive. And there is no need to be on your side neither.
I'm not a regular here, I just thought it might be helpfull to point out some flaws I've encountered while using your app. To be honest with you, I don't care if you change/fix those, as I only used AutoGKnot once, because I wanted to try it and see how it compares to the classic GordianKnot.
So once again, sorry for trying to be helpful and pointing out flaws in your oh-so-perfect app.
Gambit

[Sharktooth]

Well, joint stereo is actually the best choice.
Using J-Stereo you can effectively reduce the audio bitrate and free more space for the video encoding.

[Sharktooth]

Uhm, coz then quality would be too low...
However AutoGK main use is for backing up movies (at least i think) and J-Stereo is enaugh for that use...

[Sharktooth]

That's right. I was talking about obtaining a better sound at lower bitrates.
For example J-Stereo @ 96Kbit should sound better than Dual Channel @ 96Kbit or not?
So maybe who doesnt like 96kbit with dual channel audio could choose 96kbit with J-Stereo and still squeeze some bits for the video stream.

[guldukat]

Originally posted by len0x
It depends for what - I can tell you that on music sources I can easily detect jstereo which is quite inferior to the dual channel mode(given enough bitrate, i.e. starting with 128kbps). For movies, I don't know - if you want to reduce mp3 size why not use less bitrate?


HAHAHAHA you spoony bard, i guess you dont even know what joint stereo is, or does, right?
well let me give you a hint: "Joint Stereo: The Myths... and The Realities" (http://harmsy.freeuk.com/mostync/)

[guldukat]

Originally posted by len0x

I'll be frank that its been 2-3 years at least since I was doing listening tests myself but I could hear the difference when using lame(I guess implementation is not perfect).

well blaming an encoder implementation isnt really the way to go, besides every other modern codec uses joint stereo, and you cant even disable the use, like in vorbis for example

[jggimi]

It is perfectly fine to share useful information, guldukat, as long as you do so in a socially acceptable manner. You failed with this attempt.I find that when you describe the author of AGK as a feebly sentimental lyric poet, I must berate you for your churlish comment.

Please, reread Rule #4 (http://forum.doom9.org/forum-rules.htm).I do this even though I found myself mildly amused at your use of terminology derived, seemingly, from an English translation of a Final Fantasy video game.

[guldukat]

OT: well sorry for that, didnt know people would take that as an insult (hmmm did i forget my smilies again? oh yes i did, damn)

[Mnl]

I would like to second Sharktooth on his opinion on JS. I just went to see what they write about the topic over at Hydrogenaudio, and in the FAQ about Mp3, there are 3 different links to explanations why one should use JS:

http://www.hydrogenaudio.org/forums/index.php?showtopic=7516&hl=joint+stereo#entry74068

Also a bit farther down the list there's an explanation why abr should be chosen over vbr at bitrates below 160 kbps.

[Gambit]

Originally posted by len0x
Let me be honest with you as well then - it really bugs me that ppl who don't use AutoGK make suggestions (no matter what they are) and then get upset when I ignore them...
And I say it again, since you obviously didn't listen the first time: I WAS TRYING TO BE HELPFUL!

Originally posted by len0x
It depends for what - I can tell you that on music sources I can easily detect jstereo which is quite inferior to the dual channel mode(given enough bitrate, i.e. starting with 128kbps).
And I chalenge you to prove this in an ABX test. You know what that is, right? If not, maybe you should start visiting sites like Hydrogenaudio (http://www.hydrogenaudio.org), I think you could learn a few things there.

Originally posted by len0x
For movies, I don't know - if you want to reduce mp3 size why not use less bitrate?
Maybe because we want to reduce the SIZE while not compromising the QUALITY? And that requires to use the best possible/optimal settings.

Originally posted by len0x
But aren't both CBR and ABR modes always having target bitrate set so that final size is always the same irrespectively of which stereo mode you're using? (and only VBR mode can take advantage of that, but not in AutoGK)
Of course, with ABR and CBR you specify the BITRATE. With VBR you specify QUALITY.

Originally posted by len0x
IMHO for non-music content at low bitrates you'll be unable to really hear the difference between stereo modes (mostly because resampling and low pass filters are making too much difference comparing to the original). If you encode music then you shouldn't be using low bitrates anyway...
Ahem, the audible differences between mp3 files and the original are caused by the encoding artifacts and not by resampling and lowpassing. A properly resampled file (eg. from 48kHz to 44kHz) is undistinguishable from the original. And as for lowpassing: most people can't even hear above 16kHz. And, for example, LAME --preset 128 uses a lowpass of 17.5.

Originally posted by len0x
I'd be only happy if MP3 didn't give me the option of using JS at all - then we would not have this discussion :)
Yeah, right, why should we care about people who actually know what dual channel is good for. Screw them bilingual bards...

[Skuto]

- LAME does *not* enable dual channel mode when the -h mode is used, I have no idea where you got that idea. LAME actually never enables dual channel mode unless explicitly forced to do so by the user.

- Dual channel mode prevents the encode from exploiting *any* redundancy between the channels, *and* from allocating bits as is needed between them. It is by far the worst encoding mode, and should *never under any circumstances* be used. (This doesn't even have anything to do with the Stereo/Joint Stereo and/or stereo seperation debate. Even Forced Stereo mode allows bit allocation as needed, but dual channel does not.)

- Recommended mode (from LAME developers, and large-scale tests) for 128kbps is Joint Stereo, not Forced Stereo or (god forbid) Dual Channel.

PS. Another bug: When you use auto mode for audio selection, and for example, use a DTS track of 900M for a 1/4 DVD encode, you will end up with a 100M video stream. Somewhat more intelligent behaviour would be nice, i.e. switching to mp3 in this case.

[Skuto]

BeSweet v1.5b29 by DSPguru.
--------------------------
Using lame_enc.dll v1.32 (26/7/2004), Engine 3.96 <http://www.mp3dev.org/>.

Logging start : 12/18/04 , 17:57:14.

C:\PROGRA~1\AutoGK\BeSweet\BeSweet.exe -core( -input F:\DVD\agk_tmp\Scarface_XviD DTS T02 DELAY 0ms.wav -output F:\DVD\agk_tmp\Scarface_XviD DTS T02 DELAY 0ms.mp3 -logfile F:\DVD\agk_tmp\Scarface_XviD DTS T02 DELAY 0ms.log ) -ota( -hybridgain ) -azid( -L -3db -c normal ) -lame( -h -alt-preset 128 ) -profile( AutoGK )

[00:00:00:000] +------- BeSweet -----
[00:00:00:000] | Input : F:\DVD\agk_tmp\Scarface_XviD DTS T02 DELAY 0ms.wav
[00:00:00:000] | Output: F:\DVD\agk_tmp\Scarface_XviD DTS T02 DELAY 0ms.mp3
[00:00:00:000] | Floating-Point Process: Yes
[00:00:00:000] | PostGain normalize to : 0.97
[00:00:00:000] | Source Sample-Rate: 48.0KHz
[00:00:00:000] +-------- LAME -------
[00:00:00:000] | 'abr 128' preset is used
[00:00:00:000] +---------------------
[02:42:56:971] Gain of 1.5dB had been asserted to file.
[02:42:56:971] Conversion Completed !
[02:42:56:971] Actual Avg. Bitrate : 122kbps
[00:12:28:000] <-- Transcoding Duration

Logging ends : 12/18/04 , 18:09:42.


So AutoGordianKnot does not ask for Dual Channel. The question is then, why does it get it? :-P
Is this a BeSweet bug?

[whcanilang]

I remember I looked at my resulting .mp3's with Encspot and remembered Encspot said Dual Channel. I've trashed my logs, but I'll reconfirm on my next rips.

[Skuto]

Yes, AutoGknot certainly produces dual channel mp3's. foobar2000 can also verify this.

[Gambit]

Originally posted by len0x
Btw, guys - I was trying to remember why would I explicitly put dual channel stereo into mp3 (I cannot recall any sort of discussions about it before) since I personally do not care about mp3 audio tracks and never use them. So I checked what kind of options is used by AutoGK and it turned out that I was right - I do not set mp3 to dual channel, lame does that automatically when "-h" option is used. For more than a year no one complained - that's why its still there (and also this proves the fact target audience does not really care that much about audio part).
As Skuto already pointed out, the -h switch has nothing to do with the stereo modes, it's a quality switch (-h is the same as -q2).
And couldn't it be, that the reason why nobody complained about that earlier, is the fact that AutoGKnot is targeted at newbies, who don't want to mess around with too many options, because they don't know what they do?

Originally posted by Skuto
Yes, AutoGknot certainly produces dual channel mp3's. foobar2000 can also verify this.
And the file has a different duration. But as Skuto said, this might be a BeSweet bug in the end.

Originally posted by len0x
It wasn't even close. Guys like Mnl & guldukat are trying to help by providing useful links. I'm technical person and don't need crap "this is wrong" without any explanation.
I think I was as polite and explanatory as necessary. Why do you take my comments as personal attacks against you? I've got nothing but respect for your work on GKnot and AutoGKnot and I think AutoGKnot is a great easy-to-use app for newbies or people that want to keep it simple and effective. My only grudge with it is the sub-optimal audio encoding part.
I'm just trying to have a conversation here and my only goal is to eventually make AutoGKnot better...

Originally posted by len0x
Be warned: any follow up will be considered OT and result in sanctions described in the forum rules.
What? Do you want to ban me? For what? For finding a bug? For being more knowledgable about audio encoding than a mod? Or just for having an opinion? (I'm from a post-communist country you know, so I feel a bit of a deja vu here...)

[Skuto]

Upgrading (the included) BeSweet to 1.4 fixes the dual channel bug.

[BigDid]

Originally posted by Skuto
Upgrading (the included) BeSweet to 1.4 fixes the dual channel bug.

Hi Skuto,

You mean down/side/grading from 1.5.b29 beta(included in AGK) to 1.4 stable?

I am no expert, any chance someones report this behavior/bug to DSP Guru?

Did

[Skuto]

Originally posted by BigDid
Hi Skuto,

You mean down/side/grading from 1.5.b29 beta(included in AGK) to 1.4 stable?



Yes. I overwrote the included BeSweet+DLL's and now the output seems to be joint stereo as expected.

[BigDid]

Originally posted by Skuto
Yes. I overwrote the included BeSweet+DLL's and now the output seems to be joint stereo as expected.
Thanks from a newby

Did

[BigDid]

Originally posted by Skuto
Yes. I overwrote the included BeSweet+DLL's and now the output seems to be joint stereo as expected.
Not working, for 1 test, result is a 32kbps instead of 128, silence only, back to 1.5b29 :(

Did

[AMED]

This is straight from the Lame 3.96.1 full command line switch reference.


Joint-stereo is the default mode for stereo files with VBR when -V is more than 4 or fixed bitrates of 160kbs or less. At higher fixed bitrates or higher VBR settings, the default is stereo.

stereo
In this mode, the encoder makes no use of potentially existing correlations between the two input channels. It can, however, negotiate the bit demand between both channel, i.e. give one channel more bits if the other contains silence or needs less bits because of a lower complexity.

joint stereo
In this mode, the encoder will make use of a correlation between both channels. The signal will be matrixed into a sum ("mid"), computed by L+R, and difference ("side") signal, computed by L-R, and more bits are allocated to the mid channel.
This will effectively increase the bandwidth if the signal does not have too much stereo separation, thus giving a significant gain in encoding quality.

Using mid/side stereo inappropriately can result in audible compression artifacts. To much switching between mid/side and regular stereo can also sound bad. To determine when to switch to mid/side stereo, LAME uses a much more sophisticated algorithm than that described in the ISO documentation, and thus is safe to use in joint stereo mode.

forced joint stereo
This mode will force MS joint stereo on all frames. It's slightly faster than joint stereo, but it should be used only if you are sure that every frame of the input file has very little stereo separation.

dual channels
In this mode, the 2 channels will be totally indenpendently encoded. Each channel will have exactly half of the bitrate. This mode is designed for applications like dual languages encoding (ex: English in one channel and French in the other). Using this encoding mode for regular stereo files will result in a lower quality encoding.

mono
The input will be encoded as a mono signal. If it was a stereo signal, it will be downsampled to mono. The downmix is calculated as the sum of the left and right channel, attenuated by 6 dB.


maybe a .jointstereo file could be implemented, and for it to be placed in the AutoGK folder to enable joint stereo.

I'm quite shocked that the audio is getting encoded in dual channel.

[BigDid]

Originally posted by BigDid
Not working, for 1 test, result is a 32kbps instead of 128, silence only, back to 1.5b29 :( ...
Working now :D

Audio encoding for dummy (me)

1/ put aside the lame.dll (3.96)
2/ replace the +entire BeSweet folder(1.5B29)+ by the entire 1.4 folder.
3/ put back the newer lame.dll (3.96)

You have now access to:
- mp3 encoding in joint stereo
- also including the surround mode (dplII)
- quicker encoding by lame 3.96 ( about 20 to30% quicker)

Funny thing is it was working that way not so long ago, I made an encode of Nemo at the end of August it was joint stereo and dplII recognized as so on my home cinema decoder (it's christmas :)) I believe it was AGK 1.51 or 1.53.

Enjoy

Did

[swaaye]

I'd like to know why you people don't just stop screwing with settings you don't understand. Use the damned alt-presets. People in-the-know spent A LOT of time developing them.
http://www.hydrogenaudio.org/forums/index.php?showtopic=28124
Joint Stereo Is Better Than Dual Channel. Period.

BTW, LAME 3.90.3 is the recommended version of LAME. Anything new potentially has bugs, and has definitely not been as thoroughly tested as 3.90.3. Of course, if you just have to have the extra speed......3.96 is faster usually.
http://www.hydrogenaudio.org/forums/index.php?showtopic=28123

[BigDid]

Welcome to the forum swaaye

Being the last contributor I will answer at least to know if I am involved :D
Originally posted by swaaye
I'd like to know why you people don't just stop screwing with settings you don't understand. Use the damned alt-presets. People in-the-know spent A LOT of time developing them.
Maybe you could be more specific about "you people", there has been around 12 contributors to this thread; who are you incriminating?
Do you use AutoGK at all? If so you could know that there are no presets, no settings; choices are between ac3/DTS or MP3 abr-cbr and the bitrate, period.
Joint Stereo Is Better Than Dual Channel. Period.
If you re-read the thread it is indeed the conclusion and it is taken into consideration by Lenox with the new AGK 1.85: "
- removed high quality switch for lame mp3 encoding to get joint stereo output from BeSweet " here:http://forum.doom9.org/showthread.php?s=&threadid=64266
BTW, LAME 3.90.3 is the recommended version of LAME. Anything new potentially has bugs, and has definitely not been as thoroughly tested as 3.90.3. Of course, if you just have to have the extra speed......3.96 is faster usually.
http://www.hydrogenaudio.org/forums/index.php?showtopic=28123
This is informative thank you

Taking the time to evaluate the impact of a post +before+ posting is always a good thing, use -preview reply- for that :)

Happy new year to you

Did

[whcanilang]

http://www.hydrogenaudio.org/forums/index.php?showtopic=29422

While I've changed my preference for mp3 listening on my commute to the now preferred, listener-tested, well-regarded Lame 3.96.1 setting of "-V 5 --athaa-sensitivity 1", I guess I don't really notice that much of a difference in a soundtrack.

Perhaps because I haven't watched any videos with demanding soundtracks and I'm even less likely to ABX videos. However, my interests expressed in my original posting were for the best quality audio in the least amt of bits to allow for better video quality in the least amt of time. Except for the video requirement, those are the same factors for the Lame settings I choose when ripping cds for my little flash mp3 player. Also, since I'm ripping myself, I'd rather get decent quality without waiting around for a slower codec.

Thanks everyone for your input.

[fewtch]

Did anyone mention this yet: Lame is not tuned at all for 48 KHz bitrate. For best quality, BeSweet should be set to downconvert to 44.1 KHz before MP3 encoding step. I don't think AutoGK is doing this.

[chilled]

as far as I know, autogk (via besweet I guess) downsamples audio depending on the mp3 bitrate you choose: 80kbps downsamples the audio to 32khz and 112kbps (e.g.) downsamples to 44.1khz. dont remember the other behaviours (guess 128kbps preserves 48khz)

IMHO 112kbps (abr) is a far enough value for the soundtrack of any movie (excepting musicals and old non-spoken ones w/ background music) and then you get the 44.1 value

[fewtch]

Originally posted by chilled
[B]as far as I know, autogk (via besweet I guess) downsamples audio depending on the mp3 bitrate you choose: 80kbps downsamples the audio to 32khz and 112kbps (e.g.) downsamples to 44.1khz. dont remember the other behaviours (guess 128kbps preserves 48khz)
If so, IMO this behaviour makes no sense. What determines the filesize of an MP3 is bitrate, not sampling rate. AFAIK using 32 KHz will just produce crummier sounding audio without affecting file size (I could be wrong... never done anything that weird with my own encoding :p).

Anyway, Lame (in particular the --alt-presets) should always be fed 44.1KHz source material. All tuning/testing of the --alt-presets was done at 44.1KHz.

[len0x]

http://forum.doom9.org/showthread.php?s=&threadid=60312&highlight=LAME+automatically+downsample

[chilled]

ok thanks for the link.
so in AGK when you choose abr 96 you get a crappy 32khz sound and when you choose abr 112 you get the desired 44.1khz. In my rips with agk 32khz sound is much much worse than any 44.1khz bitrate. I guess 80kbps at 44.1 sounds better (less "metalic") than, say, 128kbps at 32khz...

so anything below 112kbps in AGK is definitely not recommended (unless len0x implements unit-by-unit customisable option in ABR mode: in this case we could choose 103kbps abr, which would mantain 44.1khz)

[len0x]

Lame 3.96.1 has better resampling decision, so at least 96kbps is encoded as 44KHz. I took advantage of that in AutoGK version 1.87.