Re: Q: two-pass quality encoding

On Mon, 15 Dec 2003, Picasso Pablo wrote:
> >> I dont use the two-pass encoding, because AFAIK it tries to keep the
> >> bitrate fluent - which I dont care about. 
> > It tries to keep _quality_ constant.
> But it does look so from my point of view. I give an example: I tried
> to encode a dvd rip without giving the codec a possibility to be
> saturated - a 17 minutes scene into a 90 MB file (720×560 or something
> like that). When I used two pass encoding (quantizers 2-31, all the
> stuff like lumi, GMC, Trellis, ... set, BVOP at 1, 1.5, 0.75) to fit
> this 90 MB, it looked awful in parts, where everything was moving fast
> - it was VERY blocky. But when I tried to fix a quantizer value to get
> the desired size (using method try-fail), the encoded sequence looked
> much better, the difference in "blockiness" between still scenes and
> fast moving scenes was minimal, thus the result looked IMHO much
> better.

Yes, a constant quantizer often give good overall quality. The problem is
just that a constant quantizer won't reach the target total filesize,
usually not even close und less (as you describe) you try several times. 
In theory, twopass should have the same effect, only a little better.
In practice, it sometimes fails...

> The second question: How do I calculate the average quantizer?

You print out all quantizers chosen in the first pass, add them together
and divide by the number of frames. How you print out the first pass
quantizers with Windows I don't know. In Linux, e.g. mplayer does the job
very well. 

gruel