On Sat, 10 Jan 2004, George Kola wrote: > > coredump or declare the file to be somehow broken? > > It was saying avi file read error.
That sounds like it may be a bug in smilutils avi I/O routines.
> > I'm curious - how was the playing time measured - with 'mplayer'?
>
> I transferred the mpeg file to windows and played it with
> windows media player.
I am beginning to suspect there is a bug in the windows media player
program.
> > What happens if you do something like this with the 52GB file:
> >
> > mpeg2enc -f 8 -M 2 -E -10 -2 1 -q 6 -K kvcd -o testing.m2v <
> > BIGFILE.y4m
> >
> > Is the resulting file still only ~16 minutes?
>
> Yes, it is.
Then the windows media player is at fault because you mention below
on that the .m2v file is 765MB
> > And that size is? ;)
>
> It is ~765MB.
>
> I found that the yuv generated is fine. I was able to
> sucessfully generate MPEG-2 with full 1 hour video with the following
> sequence
> smil2yuv ../test.avi 2>smil.err | mpeg2enc -f 3 -4 1 -2 1 -q6 -b 7500 -V
> 300 -P -g 6 -G 18 -I 1 -o test.m2v >mpeg2.out 2>mpeg2.err
Ok - so what we have is:
1) The Y4M file is fine
2) The output file is 765MB long (reasonable for 1HR at -q6 and -K kvcd)
3) With "mpeg2enc -f 8" the windows media player says the video is
only 16 minutes
4) with "mpeg2enc -f 3" (generic mpeg-2) windows media player says the
video is 1 hr long.
I am curious if mplayer says about the first .m2v file.
Sure looks like a windows media player problem with "DVD like"
mpeg-2 files.
> I did the current round on a dual xeon 2.4 Ghz box 1 GB RAM running RedHat 9.
Thanks for the info - sounds similar to my dual 2.2GHz xeon system,
but that has Suse 9.0 on it at the moment.
> The videos are for archival purpose. The researchers using
> them would be playing it only with a software player on a computer. The OS
Ah, ok - then generic MPEG-2 is fine.
> MPEG-2 and two bitrates of MPEG-4. MPEG-2 is the best quality here and
> MPEG-4 is at two broadband rates. I would like to know what is the best
What rates (and resolutions) did you have in mind for the MPEG-4
files? 320x240 perhaps? At that size ~300kbit/s is marginal
(watchable) quality but ~500 produces surprisingly watchable movies.
I'll attach one of the scripts I use with mencoder to do a 2 pass
encoding of a DV capture from a laserdisc - the max rate is set to
1000 but the actual bitrate used is less than that. You can Use
that as a starting point. It does use a couple programs from
mpeg4ip and the faac encoder to create a quicktime player compatible
.mp4 file (works when I take the file over to my Powerbook).
> flags to pass for MPEG-1 and MPEG-2 and MPEG-4 encoding (I use mencoder). I
For computer playback (not VCD) you can use VBR MPEG-1. What size
of image is going to use MPEG-1? 352x240?
> am not that concerned about the computation time. We want the best quality
> at reasonable size. We would like to keep the MPEG-1 at around 650 MB and
> MPEG-2 at less than 2 GB. The orginal video is 1 hour in duration.
2GB is not over generous for 1hr - that is about 4700 kbits/sec. To
me 'archival quality' is double that ;)
The mpeg2enc command given earlier is quite reasonable.
mpeg2enc -f 3 -4 1 -2 1 -q6 -b 7500 -V 300 -P -g 6 -G 18 -I 1 -o test.m2v ...
I would modify that to be
mpeg2enc -f 3 -4 1 -2 1 -q 5 -b 7500 -V 300 -P -K tmpgenc -E -8 -g 6 -G 18 -I 1 -o
test.m2v ...
try -q 5, and add the -E -8 and use the tmpgenc tables.
Happy Encoding!
Steven Schultz
rr.sh
Description: mencoder script
