WHAT
Lead Lighting Tech. Director Jerome Platteax and Sr. Programmer, Graphics Wright join Alexander to talk through and demonstrate the new lighting tech used in creating the Paragon trailers and on their way to you all in the upcoming 4.11 release.
Also, the December #ue4jam results are in! Tune in to hear the announce of the winners and watch us play the games live.
Question if I wanted to ¨dive in the code¨for the graphics and lighting part, where should I start and what would be the important parts and look at this?
Dynamic Emissive lighting - yay or nay ? (which can be emmited from any arbitrary shaped object, and still have correct look).
Will we get option to bake only static ambient occlusion to light maps ? There is node in material editor which lets us extract baked ambient occlusion, which can be useful for texturing, but otherwise one would want to have fully dynamic lighting and just use baked AO for materials.
New lighting tech… Please let it be a good real-time lighting solution, or an improvement of those that already exist and have not been developed due to refocus on performance.
Have thethree next steps in lightmapping outlined last year made it into the engine? Mismatch in neighboring texels has remained a problem.
Have there been any improvements to SSR, either to reduce the worst artifacts (missing reflections at the sides of the screen/around characters) or to add new features, like the contact hardening and elongation which was shown by Dice at Siggraph this year?
Is there any good method to exclude bright light sources from sky reflection capture, to avoid the double-sun problem?
Have reflection capture actors been fixed to return accurate luminance values without fiddling with Brightness?
Do box reflection captures still return invalid results when there is nothing in/near the capture component, or can they now fall back to other capture methods(~slide 73)?
Have you looked at spherical gaussians as a possible improvement for materials which are somewhat rough? Is this (or something similar) likely to make it into the engine?
Has internal research into dynamic GI been completely abandoned? Is that something which is likely to see a return in the future?
I’d also like to second questions about support for other kinds of area lights. Not having access to planar lights is a showstopper for certain kinds of arch-vis work.
Hey everyone, thanks for tuning in. I’ll attempt to answer questions that weren’t in the stream.
No progress there unfortunately. With improved Lightmass quality and build speed (Embree ray tracing) though the modular lighting seams are less of an issue - you can increase IndirectLightingQuality until neighbors match.
Not that I know of. DICE showed some good results, and Witcher 3 rendering guys had a really cool trick for solving the SSR on planar water case where the edges of the water pick up reflections off-screen.
I think we could do this easily for atmosphere since the sun disk can be controlled by the engine, but not if the sun comes from something else.
Unfortunately no, this is a nasty one. Sky lights and reflection captures are normalized separately, which is a bug.
Reflection captures are supposed to capture to infinity, and then reproject onto the bounding shape. I’m not aware of a problem if there’s nothing in the shape. There will be a problem however if the capture point of the box is inside something.
I know Brian Karis has done quite a bit of work with spherical guassians, but no plans to use them for anything at the moment.
We’re happy with it being a separate distribution. We can’t afford the maintenance and bug fixing for a complex feature that we’re not actually using.
Paragon uses mostly baked environment lighting, so Lightmass for GI. Lots of improvements there as mentioned in the stream, including 2x faster lighting builds.
I think it would be better if box reflections could return ‘clear’ results, and fall back to a probe with less (or at least, different) parallax error when there’s nothing inside the capture/transition.