[XviD-devel] [Fwd: [XviD-users] Two pass encoding time savings]

Christoph Lampert chl at math.uni-bonn.de
Sat Jun 26 09:50:23 CEST 2004


On Fri, 25 Jun 2004, J. Grant wrote:
> My query is if it is possible to create an approximation of the stats
> file? (Not creating a perfect stats file, which takes 6 hours in my
> case.)  For instance only processing a one quarter size image of each
> frame would likely produce near identical overall statistics for the
> actual encode in pass2.  Obviously the stats would need to be scaled to
> apply the data to the full size stream in pass2, but this approximation
> would save a great deal off the pass1 time.
> 
> Maybe this feature is already present?  I could not spot it.  I would be
> interested to know what people think of my idea for saving time on the
> profiling of the stream to generate the graph.

There is "turbo mode", which disables many features in first pass to make 
it faster without changing the logfile too much. I haven't heard of anyone 
trying to do first pass with smaller frame size. I guess it will change 
outcome quite a lot, since if you scale down, you lose high 
frequenciy texture information and the filesize can depend quite a lot on 
that. 
But please go ahead and test it: Do a first pass with full size, and one 
with quarter size, and compare the speed and resulting logfiles, or if you 
don't know how, send them (zipped!) to the list. 

chl



More information about the XviD-devel mailing list