Sorry for the late reply. Did you find a solution? I misunderstood what you were describing. The expression method I mentioned doesn't fix that scenario, what it does is delete deep samples wherever a given channel has zero samples.
I agree, having multiple deep samples in this way is wasteful when you're just dealing with discrete opaque objects. I've run into this issue big-time with Mantra a while back but it's the same thing you describe in Nuke: lots of sample bloat in non-overlapping opaque deep images (non-volumetric) where the only transparency was from AA and motion blur, so it seems there should be the option to have only one deep sample where the surface is opaque and facing camera, since you don't want to comp it into a volume or have it intersect with another object. Things become more complicated with glancing angles but with broad surfaces facing camera, some compression could be applied. Basically I'm describing a kind of fresnel/facing ratio-based deep sample compression. It seems like it would need to be something the render engine figures out, rather than a post process. I did kind of get it working as a comp-based solution, but it led to a little bit of aliasing due to the number of dropped samples. Cheers, Michael On 2 May 2015 at 12:22, Hugo Léveillé <[email protected]> wrote: > If you just put a simple sphere with antialiasing to high and that you use > the deep sample, you'll see some random pixel sample with multiple samples > right in the middle of the sphere. And if you use multisamples with a > moving camera, then all samples are just crazy. > > Do you have a exemple of that deep expression? > 'Thanks > > Sent from my iPad > > On Apr 23, 2015, at 8:44 PM, Michael Garrett <[email protected]> > wrote: > > Ooh I didn't know about those nodes...thanks for pointing them out. > > On 22 April 2015 at 23:37, Marten Blumen <[email protected]> wrote: > >> Do you have an example where 100% opaque front pixels don't drop samples? >> Just ran a quick test and it appeared to do that automatically in Nuke 9.05 >> when using DeepToPoints post Scanline. >> >> There are hidden DeepDeOverlap, DeepOmit & Deep Clip Z nodes available >> after updating that may try to do something you're after. >> >> On 22 April 2015 at 13:29, Michael Garrett <[email protected]> wrote: >> >>> I don't have Nuke open, but is there not a "drop zero opacity samples" >>> checkbox on the ScanlineRender deep tab? >>> >>> If not, the workaround that has worked for me is to run an expression >>> with DeepExpression that pushes the location of any deep samples with zero >>> alpha behind the camera frustum, then use a DeepCrop to get rid of them. >>> Hope that helps. >>> >>> Cheers, >>> Michael >>> >>> On 21 April 2015 at 21:18, Hugo Léveillé <[email protected]> wrote: >>> >>>> Hi >>>> >>>> Is there a way to drop deep samples in renders made in nuke's scanline >>>> when front pixel is 100% opaque? Cause as soon as you add some motion >>>> blur in your scanline, deep samples get crazy and make a very huge file. >>>> >>>> thanks >>>> _______________________________________________ >>>> Nuke-users mailing list >>>> [email protected], http://forums.thefoundry.co.uk/ >>>> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users >>>> >>> >>> >>> _______________________________________________ >>> Nuke-users mailing list >>> [email protected], http://forums.thefoundry.co.uk/ >>> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users >>> >> >> >> _______________________________________________ >> Nuke-users mailing list >> [email protected], http://forums.thefoundry.co.uk/ >> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users >> > > _______________________________________________ > Nuke-users mailing list > [email protected], http://forums.thefoundry.co.uk/ > http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users > > > _______________________________________________ > Nuke-users mailing list > [email protected], http://forums.thefoundry.co.uk/ > http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users >
_______________________________________________ Nuke-users mailing list [email protected], http://forums.thefoundry.co.uk/ http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
