[XviD-devel] Second Pass, changing size instead of quantizer

[Christoph Lampert]

Hi,

I got a (I find) rather good idea, what do you think? 

Often, I know in advance how much space a stream should 
have (e.g 1CD, 2CDs) or the bitrate (for streaming), and 
I also know what quality I want (e.g. close to constant Pquant=2,
Bquant==4). But I don't know what resolution to scale to achieve 
this.

Couldn't we create a two/N-pass mode which switches _resolution_ 
as well? I mean, most users prefer a 640x372 encoding with good visual
quality over a 720x424 with blunders in high motion scenes. 
It would need some maths, but the formula should be rather straight
forward, and it doesn't have to really be fixed-quant output, only
close to it. 

That could also give XviD an advantage for Joe-Users-GUIs as
well. Click on 1 or 2 or 3CDs, click RUN and the result will look good. 

gruel

P.S. I don't know about Windows. In Linux, the encoding appliction could
get a return value of "optimal size" and let the application deal
with scaling. But we could also ask for unscaled input and rescale
internally, fast routines are available under GPL, no problem. 

P.P.S. Yes, it's _not_ for 1.0!