[MPlayer-dev-eng] 2-pass vqcomp [was: PSNR plot]

[Zoltan Hidvegi]

Rich Felker wrote:
> Giving up on 2pass sounds like nonsense. I've used 1pass mode before
> and I can assure you the results are completely unusable... It may
> work if your whole movie is roughly constant-complexity, but if you
> have any segments significantly too complex to keep the requested
> average bitrate it will look like total crap.

I agree that vbitrate 1-pass is useless, but as I said, I'm using VBR,
constant QP, e.g. for scaled (-sws 1) 1024x576 film-based (24fps) HDTV
I usually use this:

vcodec=mpeg4:psnr:mbd=2:v4mv:aspect=16/9:vmax_b_frames=1:trell:cbp:cmp=2:precmp=2:subcmp=2:vb_qoffset=0.875:vi_qoffset=0:vi_qfactor=0.78:vpass=1:vdpart:vpsize=1000:vqcomp=0.9:lmin=4.5:lmax=4.5

This gives me around 44 global PSNR for luma, 45 average
(luma+chroma), and usually around 1mbps, but of course the bitrate is
highly variable throughout the movie.

Zoli