[MPlayer-users] Re: file size falls short of 700MB for multi pass

[Jonathan Rogers]

Jason Pepas wrote:
> should this be run on the divx2pass.log file?

Yes. That file is a simple, line-oriented, space-delimited text file 
with the same fields on each line that is easy for a human or tool to 
read. The perl script simply gathers some statistics from it.

> 
> I think I may have found the info you are looking for in another way.  I ran 
> the following:
> 
> mencoder \
>         -oac copy \
>         -ovc lavc \
>         -lavcopts vcodec=mpeg4:vhq:v4mv:vqmin=2:vqmax=2 \
>         -vop scale=640:480 \
>         -ofps ${ofps} \
>         -o ${filename}.avi ${filename}
> 
> and it was the exact same filesize as before, so from this should I assume 
> that quant 2 was 100%? ie, that in this specific case, increasing bitrate 
> would not likely increase quality?
> 

It was the exact same file size as what? This time, you only did one 
pass. Specifying the same value for both vqmin and vqmax should be the 
same as specifying that same value to vqscale. What I'm not sure of is 
whether the requested bitrate has any effect if you use a constant 
quantizer. From reading the manpage, it sounds like the bitrate is 
ignored in that case.

If constant quantizer mode ignores the requested bitrate, we know what 
your problem is: you are definitely asking lavc to use more bits than it 
needs. In my experience, I haven't been able to distinguish video 
encoded with vqscale=2 from the original.

How does your output look? I'm guessing it's nearly perfect and you 
really don't need the file size to be as big as you think you do. If 
you're scaling the input frame down (it looks like you are), use a 
larger frame size to retain more of the original detail. You may even be 
able to avoid scaling entirely. Of course, the pixels won't be square 
unless you use a frame height greater than 480 (wasting pixels), so 
you'll have to use lavc's aspect setting.

I was just reading the encoding-tips.txt file again and I realized that 
for the divx2pass.log file to be useful, one can't be using constant 
quantizer mode. So, maybe the manpage's advice to use it for the first 
pass was just leading me down the primrose path the destruction (or at 
least less than optimal video). I'm going to have to start reading the 
source to satisfy my curiosity.

Jonathan Rogers