Opinions about particle lighting in modern PBR engines?

smokeLight

Hello everyone!
I’ve been in the industry for more than 10 years now, working across multiple different engines, and yet - I always struggled to make my VFX look good in all lighting scenarios.

You make a piece of smoke that looks good during a sunny day, but in the night scene - it’s pitch black. Make some nice looking sparks, and on a different scene they approach epileptic supernova level of brightness. Make some water splashes, and it’s all good - until there is a storm, and now they are barely visible. I’m exaggerating a bit - but I’m sure you know what I mean.

I saw (and tried) all sorts of hacks: add a bit of unlit glow here and there, try to reverse the eye adaptation exposure changes in the shader, expose some parameters and give them to the level artist to play with, bite the bullet and adjust VFX for every scene manually, create some sort of system to react to day/night cycle, give up and accept that in some cases the VFX will look out of place and e.t.c.

And yet, these are still, well, hacks. Clutches to try and work around how the engine naturally works.

And it seems that with more and more games going cinematic, with physics based lights, GI, some advancements in real time volumetrics, tonemapping, directors playing around with bloom, color correction and exposure all the time and e.t.c. i’m having to implement more and more clutches every time.

I guess it partially comes down to the conflicting idea of “we want everything super photoreal, PBR and all, but we still want to see sparks and muzzle flashes during the brightest of days”, and partially - to the essence of rendering translucent stuff in predominantly deferred engines in general. I’m not complaining - it’s my job, just sharing an observation.

Do you have similar problems? Have you found some solutions that don’t feel like fighting against your engine of choice? Is there something I fundamentally missed? Would like to hear your opinion!

4 Likes

Yeah, and it doesn’t help that by the time us vfx artists get hired to come in, all the post processing and everything is already locked into the absolute worst settings for VFX every time…

1 Like

Yes, that could be even worse, especially if there is already a dozen of VFX packs from the marketplace used all over, that don’t play well together.

But as far as post processing and e.t.c. - i am now interested what kind of settings were these?
I had a problem adjusting to physically based light units once - was really weird to dial up the emissive all the way into the millions, but still, my problem was more with these settings changing from scenario to scenario, not a single set being bad in isolation.

1 Like

Technically “reverse the eye adaptation exposure” isn’t really a hack so to say, but a standard practice nowadays, for UE5 with its default exposure settings at least. Still the default unlit materials lacking the reversal of eye adaptation is kinda not helping to many inexperienced users.

One kind of post processing effect which would definitely ruin many effect artists’ day is that adjusting the overall color of the scene by scene depth.

1 Like

Hmm, some of the worst ones of the top of my head :

  • Bloom threshold set to 0 with low intensity so instead of getting nice vibrant blooms you get a white wash.
  • Replacing aces tonemapper with outdated tonemapper like reinhard cause “the colors changed” as if that’s not the entire point.
  • TAA or TSR or some DLSS or some shit that works really well on static images but makes VFX ghost and wash out like crazy.

already a dozen of VFX packs from the marketplace used all over

Or worse, a bunch of vfx artists have been building effects in separate levels /he particle simulation preview window with none of the post processing of the real game applied, so they don’t even know what their effects will look like…

In fact, to bring this back to your original point. Part of the frustration with this issue is the fact that so many vfx artists don’t even test for different lighting conditions etc…

1 Like

Technically “reverse the eye adaptation exposure” isn’t really a hack so to say, but a standard practice nowadays

Eye adaptation is meant to compensate for the effects of exposure, not for actual lighting condition changes. So I think what op meant is that they attempted to hack it to also compensate for lighting changes.

1 Like

I mostly agree, but i still call it a hack, because it is fighting against how the engine works: it wants to go physically based and realistic, and that would mean that yes, if the surroundings are bright - the glowing things are gonna be less visible and vice versa. And some games even embrace it, i.e. in Tarkov you only get muzzle flashes during night time - so i’m not sure about the idea of it being the default for unlit.

The depth though - yes, that one is unsolvable without some manual work, i believe

Ah, i see. Yes, i remember having to resort to the pre-bloom sprite based glow tactics, because the bloom was so agressive.

Well, yes, that is bad - but my point is that even if the tests do happen - i haven’t been able to make the VFX look good everywhere. It’s too bright in one scene, and too dark in another, and not even consistant in that: it may be like that for one smoke, and the polar opposite for the other. It’s like the wolf, goat, and cabbage problem - but with way more variables. The adjustments often become mutually exclusive

I remember the Source Engine had a clever solution for particle lighting without needing to do expensive light calculations- instead the engine would sample the RGB value of a single point of light at the origin of the particle system when it first spawns, then apply that tint to the particle’s color value.

Since the engine would only tint the particle when it’s first created rather than updating the color in real time, you’d have a couple artifacts here and there (If a smoke particle blows from a bright room into a dark room, it’ll stay lit up) But it allowed you to get most of the functionality of lit particles with the inexpensiveness of unlit materials.

I’ve often wondered if it’d be possible to recreate a similar system in Unreal/Unity, since it seems fairly simple.

Not to mention that there is barely any decent documentation about it thats also understandable for beginners. Looking at the past 20-or-so years, it feels like a constant reinvention of the same wheel. Sure, sometimes said wheel now comes with padding or a new color, or its an expensive wheel hidden behind paywalls, but it should be about time there’s some publicly available “vfx bible” that isnt for teaching how to VFX, but how to understand (real-time) VFX.

Glossary | RealtimeVFX Wiki | Fandom helps of course, and one can gather information here and there, but its never… tries to make a grasp expression

1 Like

It would be great to hear from DICE guys. Since SW:Battlefront they are miles ahead in translucent lighting. Still cannot believe how good this games look today.
In my experience the main problem is whoever sets up lighting does not account translucency what so ever. They use PBR color charts at best, even for the AssetZoo levels, which supposed to be a reference lighting.
Back in the days in Crytek VFX artists did both - VFX and lighting. It was the best!

Well, does anyone know if we have https://www.artstation.com/daedalus51 on the forums, by chance? He would be the best guy to ask

This has been a long time interest to me as well. I have tried just about every hack imaginable. Back when i was working on star citizen we rewrote the entire particle lighting system. It helped a lot but it still had limitations. Part of the problem is there is no one lighting model that can cover all situations. We create everything from solid objects to liquids to volumetric effects, each has its own unique lighting model. We also cover a much wider range of brightness than any other department. Toss in art directors that want to see certain elements even when the lighting wouldnt support it in a physical setting. (big muzzle flash in sunlight? yeah, no…)

Eye adaptation and tone mapping also toss a wrench into the works. I feel like its also partially a fault of us artists. We want our effects to always look punchy and dazzling, but in real life that’s not always the case. An ember from an explosion may bloom at night but in the day time its hardly visible. Our bias is to adjust it to look the way we want in whatever lighting situation we are working in, with little thought to how it looks in other situations. (unless specifically designed for that)

I feel like until everything is unified with path tracing, we are doomed to continue faking everything. Personally i don’t think its is a terrible thing however. It has caused me to find all sorts of cool tricks and grow my knowledge, including dabbling in the realm of graphics programming.

I have a thread (which needs updating!) called improving niagara lighting. It was posted with an alt account (Figment) but i put a few tricks in there for improving things. Its mostly material based but i also posted some wizardry about getting lighting on a cloud of particles using the neighbor grid system. I want to add more things i have learned in there soon.