Headac3he: Mimic Headac3 with Besweet for KVCD

rds_correia · #1 01-30-2005, 12:28 PM

Hi guys

,
I love HeadAC3he for audio compression.
I am very much to work with it.
But nowadays BeSweet seems to have a lot more options than HeadAC3.
I use HeadAC3 for KVCD and for KDVD I don't even transcode/reencode the audio or anything

I just mux the original AC3 5.1 audio track.
But a few days ago I decided to make a couple of KVCD encodes.
Since BeSweet uses mp2enc (same engine as HeadAC3) I want to start using BeSweet instead.
But most of the guides here only explain how to use HeadAC3.
Not much info on BeSweet when it comes to MP2 encoding.
I am very much new to BeSweet GUI but I already grasped the essentials.
I know most of the needed options to use in BeSweet instead of HeadAC3 but I don't know which PsyModel to use.
HeadAC3 has PsyModels 0/1/2, that is 3 different options.
BeSweet only has 1/2. Which one should I choose?
Also, how do I set the Surround2 option (from HeadAC3) when I'm using BeSweet?
Here is a basic command line for BeSweet that I've been using for Dual Channel/128Kbit.

Code:

"C:\VIDEO\BeSweet\BeSweet.exe" -core( -input "d:\mymovie\D2V\mymovie AC3 T02 3_2ch 448Kbps DELAY -24ms.ac3" -output "d:\mymovie\D2V\mymovie AC3 T02 3_2ch 448Kbps DELAY -24ms.mp2" -logfilea "C:\VIDEO\BeSweet\GUI\BeSweet.log" ) -azid( -c light -L -3db ) -ota( -d -24 -hybridgain ) -ssrc( --rate 44100 ) -mp2enc( -m d -p 1 -b 128 -e )

All help is welcome

Feel free to post all the comments that you may find usefull.
Cheers

Dialhot · #2 01-30-2005, 03:25 PM

Do not use psycho accoustic model as it ruins dolby surround ! (I don't know how you didn't notice that yet).

For DPLII, you have to configure the mode in the azid1 tab (I guess you are using besweetGUI ?). By default the output respects DPLI mode.

Boulder · #3 01-31-2005, 01:49 AM

Some kind of psycho acoustic modelling is always used, the default is p=1 IIRC, so you can't avoid it. The folks at hydrogenaudio.org have tested that p=2 is better than p=1.

rds_correia · #4 01-31-2005, 02:57 AM

Quote:

Originally Posted by Dialhot

Do not use psycho accoustic model as it ruins dolby surround ! (I don't know how you didn't notice that yet).

That's strange

I thought that it was meant to be used.
Even because it is mentioned in some of the guides.
As a matter of fact they've always been mentioned, I mean, ever since I read those guides.
On BeSweet I have the power not to use them, but on HeadAC3 there's always one PsyModel setted.

Quote:

For DPLII, you have to configure the mode in the azid1 tab (I guess you are using besweetGUI ?). By default the output respects DPLI mode.

Yes, I am using BeSweetGUI.
Could you provide me more info on what I should set?
TIA
Cheers

Dialhot · #5 01-31-2005, 02:58 AM

Quote:

Originally Posted by Boulder

Some kind of psycho acoustic modelling is always used, the default is p=1 IIRC, so you can't avoid it. The folks at hydrogenaudio.org have tested that p=2 is better than p=1.

I'm surprised. Psycho models are to remove sounds less heard by human hear than others, and the surround channel, when matrixed in the stereo stream, is easily considered as "less heard". I clearly had problems with surround when I change the psycho model. Did they test stereo or DPL ?

@Rui
According to what Boulder tells, p=1 is the default and is probably taken in consideration in DPL (or the opposite, it takes DPL in consideration) so that sounds correct that this does not give any trouble in Headac3che.

For DPL2 : go in the "Azid1" tab, check the "Stereo" box in the "Output mode" panel and select "dplii" in the combo. Then return in the "besweet" tab and see the changes in the command line generated.

rds_correia · #6 01-31-2005, 11:24 AM

Nice

Thank you both for your recommendations

Cheers guys

Boulder · #7 01-31-2005, 11:39 AM

I don't remember whether HA concentrated on standard stereo sources or DPL ones, but I strongly assume they tried stereo ones. The folks over there are more into music than movies.

tooLAME doesn't support disabling psychoacoustic modelling (there's always one used even if you set p to zero!), I'm not sure if mp2enc does. It could be determined by analyzing streams encoded with mp2enc with p2, p1 and no p set. I don't know any decent tool for such statistic analysis though.