On Tue, Oct 14, 2008 at 09:38:18PM +0400, Andrew Savchenko wrote: > -vf pp ha:a/va:a,unsharp=l7x7:0.8:c7x7:0.8,pp=dr:a,screenshot > where unsharp is definitely most resource-expensive thing (-vo gl > is used and result is too blurry on my taste with my nvidia card). With lscale=4 vo gl does the equivalent to unsharp=l5x5:0.5 I do not really like it though, encoding artefact get too visible (I have not yet made the strength adjustable at runtime).
