The fixed whole number pixel nature of 2D imaging and the decimal heavy nature of 3D is often a source of trouble.
In the case of Blender, it is easy to generate visual artefacts from scaling and moving objects around in 3D space, causing the rasterizer trouble due to the fractional positions not mapping exactly onto the 2D pixel grid.
Example: Grease Pencil thickness inconsistency causing line boil/flickering/shimmering on camera move
I am looking for, if possible, ways to clamp/snap object properties to as many whole pixel numbers as possible before each frame is rendered, but without actually touching the attributes created by the artist, or by keyframe interpolation.
It should only be done right before rendering, moving and or scaling all objects just enough (this value is usually very tiny) to align them better, to give the rasterizer more of a chance to not make things wobble and shine.
My immediate/naive thought is to convert scales and positions to 'local' coordinates instead of world, and then nudging them by that value, and I think this would work.
But I don't really know where or how to shimmy that functionality in to even test the naive version.
Are there global 'modifiers'? Should it be done with Geometry Nodes? (Which would have to be applied to the offenders manually) Should it be done with some sort of addon script that hooks in the 'render pre'callback?
Something else?
