how hard would it be to make a physically based rendering system

>>57038380
I thought that's what radiosity which was why it was so costly to render.

>>57038403
*did

>>57038418
i think that has to do with pathtracing, which is populating a bunch of rays whereas this is just assigning parameters that you still have to create artificially for scattering and such. it's just that they're REAL parameters not just specularity or quadratic attenuation.

>>57038380
Not hard at all, that's what ray tracing mixed with all the stuff you said is. It's basically how they make movies like zootopia today.

those already exist, its not an easy task though
>subsurface scattering
>physically defined materials
>raymarched lighting/reflections
>programmatic texture generation
>etc.

check ut4 or some other shit like it

>im just speaking out of my ass

>>57038440
my idea is
-to combine pbr with other stuff like opensubdiv for dynamic tesselation and level of detail with realtime peformance in mind.
-Texture models using per-poly uv generation via ptex. Non visible uvs can be dumped from memory for overhead management in realtime testing.
-Parallax mapping for large assets such as architecture where overhead is avoided. Displacement for the rest. Each utilized under ptex UVs.
-Bullet 3 for gpu collision detection and rigid body dynamics. Possible verlet integration for hair and cloth.
-Openvdb for fluid fog and destrutive simulation outside of bullets data tree efficiency.
-Terrain with heightmaps and support for booleans for making caves
-Depth map for fog generation and possible tonemap compositing.

too much?

What's your end goal?

Sorry OP, you need to be at least able to sum 2 digit integers. just give up

>>57038507
Many of these exist/have been done before/are industry standard. What's your point?

>>57038744
those are all used separately though and in offline configurations i'm talking about realtime implementation via lowering a lot of the sampling rates.

>>57038380
Sounds pretty hard but I don't know shit.

File: 512_001.png (1MB, 2048x512px) Image search: [Google]

1MB, 2048x512px

>>57038380
I made a raytracer for uni, we had a course where we built up on our implementation as the semester went on and we learned new stuff. It was pretty interesting, but the theory became quite insane towards the end, really math-heavy. I spent whole nights debugging my code and I'm not even sure I got it exactly right at the end.

The basics are relatively simple, what's difficult is trying to get a good enough estimate of real-world light physics while tracing the minimum amount of rays to render your scene in a reasonable amount of time.

Pic related took like 15 minutes to render and it still has artifacts (all those bright white dots). And made 180 of those images to make an animation. It doesn't even use meshes, just dumb old spheres. And all textures (except the earth) are procedurally generated, the Moon and Mars even have normal mapping.

>>57038507
You have no idea what you're talking about.

>>57040367
not an argument.