I have few question as well. One that I think everybody is curious in long term:
Global Illumination. Are you guys going to work, on LPV system (which looks quite promising), or are you considering alternatives. Like Deferred Radiance (Irradiance) Volumes, Delta Radiance Transfer, or other SH probe based solutions ? Can’t pretend I understand fully how they work, but from What I understand both LPV and radiance transfer solution are viable, if enough tweaking to end user is exposed.
Any plans to introduce OptiX baking solution for static lighting ? I can sacrifice my editor usage, for 3 minutes level bakes ;).
Now important questions for me:
Will be there Forward Pass rendering for proper Order Indepedant Transparecy sorting ? And more over proper transparent surface rendering. So transparent surfaces can be 100% opaque, without bleeding ? Currently transparency is in flux. It is especially annoying when trying to use transparent Decals. In some cases bleeding is acceptable in other (when you place brick decals over plaster wall) is not. Will be there possibility to place decals directly on model within DCC app, and then export it, and apply decal material to that surface. I currently use floating geometry for it, but you can imagine, it not always fit and there can be z-fighting issues.
Foliage Lighting Mode. Will be there foliage lighting mode in foreseeable future ? Currently there is no way to make lighting condition independent, realistic looking foliage. What would be needed here is nice uniform lighting mode, for two sided materials, that is more surface normal independent.
Which also takes into account directional lighting, and ambient light.
Are there any plans at all to reintroduce custom lighting mode to engine ? I know that we need Forward Pass rendering to do it. But honestly we need it anyway. I don’t think the issues with translucency sorting can be solved in deferred pass. Though I might be wrong!
In any case reintroducing custom lighting mode would solve the problem with foliage.
Any plans to improve shadow rendering ? I mean current system is “all or nothing” within set distance. It would be more efficient to have system that takes into consideration object size/distance ratio. So small objects will cast shadows only closer to camera, while huge hero objects that might be in background (like big building), will cast shadows from far away.
Any plans to use shadow map cache for further optimizations ? Static objects do not need to update shadow map every frame. It can be updated every 4 frames for example. Useful for scenarios with time of day, where lighting is changing over time but not fast enough to make cached shadows look odd when changing.
Out of curiosity. Did you guys considered using Sparse_textures for shadow maps ? I was unable to find any paper on it, but in theory it could allow for nearly infinite quality shadows maps, with very little memory usage. It can be interesting research direction!
Will area lights be back ? By area lights I means lights that affects diffuse color from arbitrary shape. Though I will be happy with just rectangle!
Any plans to introduce real volumetric lighting ? I mean volumetric lighting that can be see scattered in fog or dust when you are not looking more or less in direction of sun.
Perhaps you can release the “current/beta/wip” high overhead version you have for us with the more powerful system and the community can help in tuning it ?
I have a niche question. I work in a lab that does a lot of academic research involving flight sims. We have a lot of work invested in aviation instrumentation/GUI components that pump out C++ OpenGL code that can then be dropped into any GL renderer assuming we have decent scene graph access. We’ve done this with many renders including our own OpenSceneGraph solutions to commercial products from the simulation world. Although it forces us to a GL only version of the engine, do you think such an integration would be difficult with the UE4 pipeline. If not, where do you think the best place to hook in our code would be?
Thanks for all the great work you’re doing; the render is really impressive. Really well balanced AA solution!
They covered that in the last Live Stream I believe, there’s no plans for it right now. Static lighting still gives you accuracy that dynamic lights can’t, and the current stationary lights are a good compromise IMO.
Custom Lighting support/support for user-created lighting models still remains my topmost question personally. Do we simply submit questions during the stream?
Actually, I’m curious about the stuff that people have already asked about: volumetric lighting, volumetric shadowing, area lights, area shadows, more detailed shadow rendering for different objects, contact-hardening shadows (although, area shadows ARE actually contact-hardening shadows by the very definition of area lighting), etc.
Nothing special, to be honest…
No they only covered why they removed SVOGI.
I’m actually really curious what other solution they are currently examining for implementation. LPV is already here, and I would love to know if they are planing on expanding it or implementing something different.
I would also love to know about GPU precomputation for. I willing to sacrifice my editor working for it, if it can reduce rendering time from 30mins+ to 5mins.
Got into a discussion last week with some rendering guys about how VR is forcing post fx algorithms and other algorithms to be implemented differently, since they assumed 2d projection. How are you guys handling this? will their be tools or perhaps material editor things to help tackle this?
Hi everyone, thanks to all who were able to make it!
Twitch automatically archives the stream and you can view it here now](Twitch). We’ll also post it to Unreal Engine - YouTube in this playlist]() most likely by Friday.
By this time tomorrow, all active UE4 subscribers will be able to download the Effects Cave showcase for free from Marketplace.
Follow up rendering notes from Martin M:
DBuffer:
To use the feature enabled put this into consolevariables.ini and [re]start the engine (or [SystemSettings] in your project ini):
r.EarlyZPassMovable=1
r.EarlyZPass=2
r.DBuffer=1
You need to change the material to use one of the DBuffer DecalBlendModes (open the details on the final block).
If you want some materials to not receive DBuffer decals you can specify that in the DecalResponse (same location).
Documentation is upcoming.
LightPropagationVolumes (LPV):
To get the an early peek at the feature enabled put this into consolevariables.ini and [re]start the engine:
r.LightPropagationVolume = 1
At the moment this console variable cannot be changed at runtime.
Initial scene setup:
Add/Change the directional light to be “Movable”
Add/Adjust geometry and the light direction to give you some shadowed parts and some parts that receive the light (ideally bright colored materials).
Make sure you don’t have any Lightmass enabled to be confused as dynamic (both techniques together would result in too bright indirect lighting).
Enable “Affect Dynamic Indirect Lighting” on the light source
Documentation is on the way.
Thanks again for all your questions and feedback. Keep watching the Announcements section for details on our next stream.
