SVCD/MP2 Audio FAQ
MPEG-1 Layer 2 & MPEG-2 Multichannel Questions
1. What formats are supported by SVCD specification?
SVCD supports two audio formats officially:
-CBR MPEG-1 Layer 2 Audio, 32-384 kbit/s, 44100 Hz sampling rate, 2 channel, up to 2 tracks
-MPEG-2 Multichannel Audio (although allowed in the spec, it is supported by very few players)
2. Is it possible to have 5.1 channel audio on my SVCD?
Yes, there are two ways. The most reliable is to use a Dolby Pro Logic or DPLII downmix when encoding to 2-channel MP2. Simply add a -surround or -surround2 to your besweet commandline in the azid section.
The other way is to create a MPEG-2 Multichannel 5.1 stream. Standalones that can decode these streams, however, are few and far between.
3. How can i convert AC3 into Mpeg2multichannel?
You would need the phillips (
encoder), and
BeSweet. Here are the two commandlines :
Code:
BeSweet -core( -input input.ac3 -output inter.aif -6chaiff ) -ssrc( --rate 44100 ) -ota( -g max )
pub_enc -l 2 -m s -n d -r 44.1 -b 384 -L -p 3 -e -v 0 inter.aif final
The resultant file will be
final.mpg which is a mpeg2multichannel stream.
[If anyone have the phillips encoder maybe we can put a new link, BTW I think than nobody need that old surround method not fully supported]
4. What MP2 encoders are available?
The two most common encoders (both included with besweet) are
TooLame and MP2Enc. Both are open source, and MP2enc is the encoder from the CDex project. Although extensive testing has not been done, both provide good and similar quality.
[Of course ffmpeg can encode to .mp2]
5. Is it possible to increase the volume or normalize MP2 without decoding & reencoding (i.e. losslessly)?
Yes, the only tool that offers this feature currently is BeSplit. An example command line is
Code:
BeSplit -core( -input infile.mp2 -output outfile.mp2 -type mp2 ) -ota( -G max )
6. Which psymodel should I use with toolame or MP2enc?
There is much debate about both the psymodels, and the only fully correct answer is to use your own ears and judge for yourself.
In both toolame and mp2enc, Psychoacoustic Models 1 and 2 are the ISO standard models. Although they are standard, they are not to be considered very high quality. However the psymodel does not have a large impact on the sound quality. In general, Psymodel 2 is considered to be of better quality, but may have problems with frequency cutoff and tonality detection. A guideline used by many users is to use psymodel 2 at bitrates less than or equal to 160 and psymodel 1 at higher bitrates.
Later versions of toolame have implemented several new psychoacoustic models: psy 3 is a re-implementation of the standard psy 1 and psy 4 is a cleaned up version of psy 2. Although these versions will not sound much different from their ISO standard counterparts at the moment, there is possibility of their improvement in the future. See
this page for more information on toolame's psymodels.
7. Should I use stereo or joint stereo for MP2?
In general it is recommended to use stereo for MP2, the joint stereo in MP2 is intensity stereo only, not M/S (mid/side joint stereo as used in lame and many other codecs), so joint stereo is probably a bad idea except at very low bitrates (less than 128kbps).
8. I have a dual channel MP2 with a different language in each channel. How can I separate the languages?
BeSplit v0.9b6 and up can accomplish this losslessly by producing two mono .mp2 streams. Note that this is only possible if your mp2 is
dual channel. Use besplit from commandline:
Code:
BeSplit -core( -input stereo.mp2 -prefix e:\channel -type mp2 -demux )