On Sunday 26 January 2003 11:20, Bernhard Praschinger wrote:
>Ok. But currently there is no way to contol the size of the output file.
>On a SVCD with SVCD settings fit usually about 40-60 Minutes.
>My personal workaround is spliting the video with editlists into some
>parts, and try to fit them as base a possible on the disk. So I may only
>have to recencode one or two part, so the they fit nice on the disk(s).
> You should set a much lower quality factor, than you might get a better
> VBR stream.
> When you take the look at the howto, look at the sections: Creating
> MPEG2 Videos, subsection: Which values should be used for VBR Encoding.
>
> auf hoffentlich bald,
>
> Berni the Chaos of Woodquarter
Thanks that's very helpful. Since I've been seeking to trade size against
quality and speed wasn't so critical I had assumed that a low quantisation
value would produce better results independant of bit rate. Given I usualy
try to fit too much on a disk I reckoned having a very low -q was good.
Re-reading the FAQ, which seems to have been nicely updated since I last
looked, has clarified that a lot.
My current piece of video that I'm working on is recorded vhs captured as dv
files and a value of -q 8 certainly seems better and faster. Reading the FAQ
again also suggested to me that using -l 0 would speed things up and it does.
Many thanks,
Michael.
--------------------------------------------------------------------------------------------------
--------------------------------------------------------------------------------------------------
As a footnote below is how I have been doing my calculations:
i). First I calulate the available space in kilobytes on my media:
Number of CD's multiplied by space on CD *1024
e.g. for a single 80 minute CD I'm using 795*1*1024.
So that is 814080 kilobytes.
ii). Then I calculate the number of seconds of video
hours*3600+minutes*60+seconds (now I'm pretty confident this ones right !!!!)
e.g. for 50 minutes its 3000 seconds
iii). Next the bit rate that will fully occupy available space
8*(available space/seconds duration)
e.g. 8*(814080/3000) equals 2170.88 kbps
iii). Finally I subtract the audio bit rate to determine how many kbps remain
for video
e.g. 2170.88-224=1946.88
Rounded down that gives me 1946 kbps desired average video bit rate.
Given that (I believe and may be not quite correct) one sets the peak rate
with -b then setting at around 15-25% higher seems sensible.
I guess I'll have to accept that for vbr streams this just gives me a starting
point and continue with the human multi-pass system:
BEGIN
roughcalulation;
REPEAT
transcode ; multiplex;
guesstimate_new_value;
UNTIL good enough;
END
:~).
----------------------------------------------------------------------------------------------
Also, in case anyone decides to try my script I felt I'd better repost with
corrections. Don't use what I originally posted - you'll probably get
terrible audio-video out of sync. I cut and pasted the version I was playing
around with. My current production script (improved thanks to Bernhard's
advice about -q) is (again apologies for word wrap) :
#!/usr/bin/perl -w
use strict;
my $smilfile;
while($smilfile = shift()) {
if($smilfile !~ m/\.smil$/) {
print "This is not a .smil file! [$smilfile]\n";
next;
}
my $audiofile = $smilfile;
$audiofile =~ s/\.smil$/.mp2/g;
my $videofile = $smilfile;
$videofile =~ s/\.smil$/.m2v/g;
my $mpegfile = $smilfile;
$mpegfile =~ s/\.smil$//g;
print "Video transcoding : [$videofile]\n";
system("smil2yuv -i raw -a \"$audiofile\" \"$smilfile\" | yuvdenoise -F -f
| yuvscaler -v O -O SVCD -n p | mpeg2enc -v 0 -a 2 -f 5 -b 2200 -q 8 -4 2 -2
1 -I 0 -S 795 -B 224 -N -V 230 -o \"$videofile\"");
print "Multiplexing [$mpegfile]\n";
system("mplex -v 0 -f 5 -m 2 -b 230 \"$audiofile\" \"$videofile\" -o
\"$mpegfile%02d\".mpg");
print "Done\n";
}
-------------------------------------------------------
This SF.NET email is sponsored by:
SourceForge Enterprise Edition + IBM + LinuxWorld = Something 2 See!
http://www.vasoftware.com
_______________________________________________
Mjpeg-users mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/mjpeg-users
