[MPlayer-users] 2 pass with lavc, 2nd pass looks a little worse (only a little)

[D Richard Felker III]

On Thu, Aug 14, 2003 at 12:36:17AM -0700, upgrdman at mindspring.com wrote:
> [Automatic answer: RTFM (read DOCS, FAQ), also read DOCS/bugreports.html]
> in the docs, it says that the second pass should look considerably better than the first pass, but for me it doesnt. it look the same it some parts, and WORSE in others. i dont think this happened before, but i cant remember the version of mplayer/mencoder that i used to use that worked great. it was an old old version, like 0.60, or somewhere close.

Please use a mailer that wraps lines, or press the enter key somtime!
Otherwise it's hard to read stuff like the above!! I've rewrapped the
rest of your email in the reply.

> now my second pass looks, like i said, the same, with some parts
> looking worse. i looked closely... even on scenes where it looked
> bad on the first pass, looked jsut as bad on the second pass...

This is very strange. Is there anything special we should know about
your input file? (lots of noise? lots of flashing or rapid motion?) A
bitrate of 1000 should be more than a plenty for great quality with
most movies. In fact, I would recommend not using such drastic
single-coefficient elimination (v*elim), since it will lower the
quality.

> im compressing a dvd-rip, ntsc, and according to mencoder, its
> telecined, so i set the ofps to 24 :) and it duplicatres LOT less
> frames than it used to :) (but this is not what appears to cause my
> problem with mediocre 2nd pass outout fules.) and yes i know
> technically it should be 23.973 or something like that, but 24 is
> good enough for me :)

If you use 24, you'll get a duplicate frame about once every 40
seconds, making video look choppy (especially during smooth pans). Why
do you insist on using the wrong -ofps? BTW, it's supposed to be
23.976.

> mencoder NationalSecurity.vob -aid 128 -oac mp3lame -lameopts abr:br=96 -ovc lavc -lavcopts vcodec=mpeg4:vbitrate=1000:v4mv:qpel:vlelim=-4:vcelim=9:lumi_mask=0.05:dark_mask=0.01:vhq:vpass=1 -sws 2 -vop scale=608:336,crop=720:466:0:4 -ofps 24 -o NS-compressed-1.avi && mencoder NationalSecurity.vob -aid 128 -oac mp3lame -lameopts abr:br=96 -ovc lavc -lavcopts vcodec=mpeg4:vbitrate=1000:v4mv:qpel:vlelim=-4:vcelim=9:lumi_mask=0.05:dark_mask=0.01:vhq:vpass=2 -sws 2 -vop scale=608:336,crop=720:466:0:4 -ofps 24 -o NS-compressed-2.avi
> 
> (it should be all one line, but i think my email client messed it up...)

Nope, your email client is too dumb to wrap lines... :)

> so any ideas? if necessary i could upload the first and second pass
> output files, but they each ~700MB... which kinda surprises me
> actually... theyre so close in size to each other...i think less
> than 1MB difference. shouldnt the second pass have a bigger or
> smaller gap in size?
> 
> any ideas?

Eh? BTW, the only thing that would be useful for testing is the
*source* file, and I imagine that's way too big to upload. My guess is
that there's one VERY intense scene somewhere, that's taking all the
bits in the second pass, and still not having enough to look good, so
it's making everything else look bad. You might just have to use
vrc_override to override quantizers for that scene....

Rich