Please see https://lowrez.marklowne.com
USE IT AT YOUR OWN RISK!
These plugins allow fast (really! heavy algorithms are hand-optimized) 'true 3D' relighting of rendered sequences, given a depth channel and the 3D camera. This means the lighting will take into account the actual position of the surfaces and not only their normals. And this means, in turn, that you can relight using a spot light with quadratic decay, for example.
Now, here's why 'true 3D' is NOT true 3D (and you can't kiss mental ray goodbye - yet):
- no REAL shadows - unless you provide a real (hopefully low-poly) 3D model of the scene
- no raytracing - no reflections, no refractions, no nothing
- no light volumes of any kind
There is a video showing Relight in action in the plugin's page at my site, check it out!
v0.7.1 - Linux is back
- somehow ported to Linux again - dynamic_cast<> and vague linkage aren't really cooperating, workaround is not very elegant...
- rewrote internal hash management, now relight3D should work correctly with DiskBuffer (v0.2.0) ;)
- fixed 'external' shadowmap update bug
- fixed silly alpha-related bug in bump3D
- fixed seldom-occurring requested channels bug in rlMaterial
- added support for geometry-based shadow maps in rlLight, and rlShadowmap gizmo to produce them effortlessly (a dirty hack, but works)
- added support for light projection in rlLight
- added lookat input in rlLight
- added support
for color, specularity and reflectivity passes in rlMaterial to spare
the lazy compers some nodes
- added position3D convenience node and the ability to use a position pass instead of depth in all nodes
- added normal3D node to compute normals - or just smooth them - from depth or position pass
- added bump3D node to bump normals via UV bumpmapping, color pass luma or 3D perlin noise
- added occlusion3D node - fake ambient occlusion, most effective with high-detail geometry
- rl3D is now called relight3D for no particular reason, the same for relightIBL
antialiasing (coverage based and/or normals based) to relight3D, with
the additional benefit of solving the interactive zoom
- faster and more
accurate algorithms, some parameters now behave differently (e.g.
- tweaks everywhere
- more minor bugfixes
- File/Buffer reports will crash...
- there could be multithreading issues
- Linux version is as untested as it could possibly be, unexpected behaviour is to be expected
- reorganize everything from scratch once again to allow for more complex materials - that is distant future...
Any testing and feedback VERY appreciated.