Download

Getting the right falloff without overlapping stationary lights

I cannot seem to get the right falloff from my lights. All are stationary, it just seems unrealistic. I have tried every solution I can find, I have literally spent the last 15 hours trying to figure this out. 15 hours of researching, building lighting, trying different things etc. I cant figure out what im doing wrong. My game has dynamic day and nig but takes place indoors, so I cant have static day or night. I tried reflection captures and those helped a little but the falloff problem persists. The only thing that helps is putting more lights, but then they overlap and i get the red X on it. Not sure if thats just something that has to happen for the scene to look good or what. Here is my scenes. Two different hallwas in the mansion, one with lights on the one wall and one with lights on both.

The second picture:

If you have a day/night cycle, do you bake anything at all? If so, you don’t even need Stationary Lights. The red x comes from 5 overlapping Stationary Lights that cast *STATIC *shadows, when the 5th light is placed the smallest one will get the X. I don’t even see any shadows in your scene, but if you need the baked indirect lighting still you can disable static shadows on the Stationary Lights. But if you don’t, it’s just easier to use Movable lights, Stationary/Movable are identical unless you’re baking.

For my day night cycle i use the Ultra Dynamic Sky Blueprint from the marketplace which has a setup that allows me to select 2 directional lights to use as a sun and moon, both of the directional lights are set to Moveable, so nothing is baked there. All 76 of the lights in the mansion are Stationary. They are in a blueprint, not sure if that matters at all or not. I use a point light in a sconce, there is no regular point light in the map outside of a blueprint. I went with stationary since the static lights look terrible when baked, and moveable has way longer render times so i figured stationary was a good middle ground. I dont really want to bake so I think i’ll change them to movable which gets rid of the red X’s, however the problem of indirect lighting still exists.

Right now im using a post process volume with Indirect lighting intensity set to 10.

The first picture is indirect lighting off, the second is it on.

The lights have been changed to movable, and since theres no baking needed and no red x’s now, the performance should be better, I checked with the Light Complexity and Light Density Views and everything now seems to be within acceptable levels. Im not sure if using the post processing to help with the light falloff is what people normally do or what but it seems like thats an issue that could be addressed in the lights settings but i havent had any luck with that, but then again i didnt try it with movable lights.

Dynamic shadows have a big impact for performance, so it will help if you turn off shadows for some of the lights

Yeah that was a big concern of mine, but the entire game will be indoors and even then only 4 or 5 lights will be on the screen at once. I set the Lights max draw distance to a distance that will show to the end of the hall and all others will turn off. On top of that since there’s no need for baking now, I can cut my number of lights from 76 to probably half of that since i can just up the intensity and attenuation radius, but we’ll see. The mansion is pretty empty for the most part. More stuff means more shadows so It could potentially be an issue down the road.

I could probably just turn off individual meshes shadow casting ability so the scene doesn’t look so weird with the light not casting any shadows if it becomes an issue. If things get bad then I guess its either back to static or stationary, or I can mix movable and static. So far I don’t have any performance loss. I get a steady 8.33ms no matter where i go. I tested in PIE and standalone. I could open the Classic Mansion Project that I got from the marketplace, which is where the assets im using in the pictures come from, and I can test changing those lights to movable and see what happens, theres thousands of objects in there so it would give me a pretty good approximation of performance in the final version of my game.

I was thinking of using raytracing. I have an RTX 2070 so I could potentially do it but I think thats even more expensive than my current lighting setup.