#as I made the probes denser, the problem
rancid flare
I just said it's baking at defaults... just gotta wait till its done
which should be any second now 😮
ornate gale
What I expect is that denser probes may make the problem that's already there appear in more places, but overall reduce the inaccuracy and leaking
For your own sake you should be baking smaller sections of a scene when testing
And smaller scenes in general when practicing
rancid flare
ive done that already, now im into the phase of implementing it in a larger environment and encountering new problems that didn't occour in the smaller test
ornate gale
A local APV volume is a good alternative over scene baking before you have the settings nailed down
ornate gale
Looks like an effect of insufficient indirect samples
rancid flare
I know the intensity of these particular lights is part of the problem
i considred putting them on a rendering layer excluding them from the probes
or baking the light effect into a decal
ornate gale
But the light is leaking less around the walls now
Before the leaking was obscured by how few probes ended up being lit by the indirect samples, relative to area
rancid flare
those red walls don't have a lightmap
they are moving objects so that light is fully from the probes
my old pre-light probe method would have been to dynamically light them from a separate light layer based on the environment
ornate gale
Yet they are contributing to GI?
rancid flare
they are stationary long enough that having the light from them baked into the map on the ground is worth while. They only move when triggered
ornate gale
So them moving doesn't really matter for the purpose of baking light
rancid flare
sometimes in other cases its been worth it to put a static lightmap on moving objects without casting a static shadow simply because their range of motion did not meaningfully impact the lighting recieved.
ornate gale
Like I said, more indirect samples to smooth out the "fireflies"
rancid flare
okay
ornate gale
More probes if you want to improve the direct shadow resolution, and to reduce light leaking around the gaps of the red walls
rancid flare
ill bake it and see
well like I said... not too worried about shadows from probes
just bounce light
ornate gale
How much is up to you of course
rancid flare
the game is going to be very stylized with a step based cell shader
but I turned all that off for this since i work in an actual studio and that would be showing too much
but the same problem exists without the shader
or with*
ornate gale
And also use the APV debugging to verify the fireflies and other problems aren't actually from probes that are supposed to be invalid
rancid flare
the hot spots as you assume are from probes touching surfaces
but the indirect samples were set quite low (128)
in previous projects we used fully mapped/dynamic light on rendering layers (basically very old school) so low samples were fine
ornate gale
APV virtual offset volume is very useful
Offsetting probes away from surfaces can even out the light they receive, in general but especially if some are inside geometry without being properly invalidated
rancid flare
example from the last project
but theres no probes at all in that game
its all lightmaps and layers
ornate gale
Like I said APVs require more samples on average
rancid flare
but yeah, im probably the opposite of most people on here XD lots of young people rusing into things. I'm an old dev trying to not be stuck in my ways lol
ornate gale
but yeah, im probably the opposite of most people on here XD lots of young peopl...
Learning new techniques is hugely important, though especially when it comes to lighting they are quite frontloaded with necessary theoretical knowledge to use them effectively
rancid flare
I have made some interesting discoveries
environmental samples seem to have a bigger impact than indirect, but indirect is involved
but... the most mileage came from simply disabling the indirect on the bad lights
they were so bright to throw across the ground that nearby probes could super-amplify a single ray and cause teh sparkle cancer
This theoretical case here I was able to completely eliminate all but the very small amount of edge wrap of the light with these settings.
but know that the red light has a 0 indirect multiplier
if i turn it to 1x it causes sparkle cancer
This image is at 1x indirect on the point light, but all other settings the same
ornate gale
but... the most mileage came from simply disabling the indirect on the bad light...
Getting rid of it is one option, but it won't fix issues with it obviously
ornate gale
environmental samples seem to have a bigger impact than indirect, but indirect i...
It's something I don't fully understand but in my experience APV baking uses both indirect and environment sample sliders for indirect samples, specifically picking the one that's bigger
Additionally Override Sample Count volume doesn't expose environment sample count at all
So that's probably why environment samples seemed to work "better"
Though in the image with 2048 environment samples visible, there is no indirect or environmental light at all to be seen
Usually when I bake APVs with very bright long range lights I use a lot more samples than you do
In particularly problematic areas I may have to use an override to increase sample count to 512 indirect with 256 multiplier
Which adds up to 131 072 indirect samples
So I wasn't kidding when I said APVs need more samples
But generally that's the worst case scenario when those bright lights are very close to high albedo surfaces
rancid flare
Ill try that tomorrow