Fredrik Pihl wrote:
> I run all and view in linear.
> And I cannot replicate the grade exactly....

3D LUTs don't tend to work very well on linear images, as you're
discovering :)  Because the values for the shadows are all squashed
into about 0.0...0.1 it's hard to get enough points in that area, even
with massive 64x64x64 cubes.

I'd try converting to Cineon for the 3D LUTing part.  The Vectorfield
node has input and output colorspace options for exactly this reason.

Under your CMSTestPattern try a Colorspace set to Cineon in, linear
out.  Then your grade nodes.  Then another Colorspace set to linear
in, Cineon out, then GenerateLUT.  Once you've made the LUT you can
use it on linear data with a Vectorfield node as long as you set its
two Colorspace dropdowns to Cineon.

This also allows you to use 3D LUTs for the full Cineon range, so
you're not limited to linear values from 0..1.  You can use something
like -0.1 to 12.0.

I admit that I still see slight differences between different
applications using the same 3D LUT sometimes.  I think this is due to
different interpolation methods.  I've never managed to get Houdini,
Flame and Nuke to agree 100% on what the same LUT should look like,
but with 32x32x32 cubes it's not enough of a difference to be a
problem.

-- 
Lewis Saunders
trilinear fields for life
London
_______________________________________________
Nuke-users mailing list
[email protected], http://forums.thefoundry.co.uk/
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users

Reply via email to