Lighting is one of the complex subjects that a lot don't get right or just don't bother to get right due to its complexity. It would be nice if Unity expanded the Unity Learn section on lighting. Using the new APV system. How to create nice lit scenes from A to Z. From smaller scenes to larger scenes. How to do proper Day / Night cycles. All for URP and HDRP.
Exactly, we end up having to scour the web for tips and tricks from different sources and attempt to compile them to work for our scene. Usually "good enough" is the level of quality we end up with lol.
Efficient? Not at all, there is no adequate runtime leak prevention tech for APVs. The offline touchup volumes don't do it enough justice in a lot of cases, because there are many scenarios with valid, visible rooms on other sides of the thin enough walls. You're just throwing artefacts from one side to the other, as even visible at 14:31. Plus, for our artists it's a hassle to control these volumes and use debuggers, manipulating the grids of probes feels easier for them. And biasing in itself introduces a lot of caveats as well. A runtime depth-based or visibility-based invalidation of probes is necessary to determine if a probe at a sampling position should be contributing. Though, understandably, that may require a lot of storage (L2 SH needs 27 floats and there are packing optimizations, and now compare that to 64 octahedral depth samples). In fact, you tried to experiment with octahedral depth and even have semi-working lightmapper output support for them, but I imagine you didn't implement it precisely because of the storage concerns, am I correct? Heck, how about just using a rotated cube depth encoding like in TLOU2? It had very sparse hand-placed probes, but APVs can have so much more density and give accurate results, so why even not? Or is it a problem because now you can't do hardware interpolation of the probes in a brick and will require 8 full loads? But other than that, it's a solid tech overall, but runtime leaking prevention is absolutely necessary in order to make it truly useful for us and, I believe, many others.
@@thomasfalcon6350 I can assure you that APV is far from being an obsolete technology, as realtime GI solutions in themselves have loads of limitations. Something like Lumen is a cool piece of tech, but you sure as heck ain't gonna run it on a last-gen conole, mobile phone or a low-end desktop GPU. And good luck fighting the noise and temporal artefacts, especially if your project cannot support mandatory TAA. And also good luck having reduced precision and artefacts for tranparencies due to them relying on Lumen data projected onto the camera-relative APV equivalent in Unreal.
@@thomasfalcon6350 Most game/realtime 3D projects are not aimed at platforms/specs that can support realtime GI at respectable frame rates in the coming... Lets say, 5 years? Likely longer if we take mobile and standalone VR into consideration. So from that perspective it's reasonable to suspect that APV will remain usefull for years to come if they work out the kinks and continue to support it. And besides that, HDRP already supports raytracing, specificaly path tracing, that once the hardware of the average joey catches up to it, already provides your desired proper realtime GI solution. So your claim that this is a "obsolete technology", i do not understand how you could say that, and i find my self on the other side of that argument. I think this is a great piece of technology that fits the bill for the current requirements untill we likely reach a point years from now, where most to all hardware can finaly provide realtime GI for all users. Untill then, things like APV are required to fill the gap.
@@thomasfalcon6350 Realtime GI is too expensive for mobile platforms. Besides, the most impressive lighting in AAA games also still depend on lightmapping and baked GI. It is far from obsolete. The fact that the APV system supports per pixel lighting makes lightmapping in many cases already not a necessity anymore and help iteration time. I do agree that in some cases a better realtime GI system would be preferable, but only good GPU's can run it. Don't be fooled by Unreal Engine tech. Things like Lumen are very impressive, but it comes with a cost.
afaik you can mix them, but you have to choose either "Receive Global Illumination: Lightmaps" or "Receive Global Illumination: Light Probes" per each MeshRenderer. At least with the latest beta version that I tried :)
I try to do this but find out my scene always baking the lightmpat, even when i press the bake button in the probe volume menu.. that's why mine took very long time to bake, didn't find the tutorial how to bake with probe volume only. And after the bake complete, i try to add sphere mesh with hdrp/lit material, then try to moving it around to try if the APV working, but the sphere still look flat all over the room not affected with all the probes, or simply it's just my wrong step.
I think you need to enable what source of lighting will affect each object, like Kai F said above " Kai F 2 weeks ago afaik you can mix them, but you have to choose either "Receive Global Illumination: Lightmaps" or "Receive Global Illumination: Light Probes" per each MeshRenderer. At least with the latest beta version that I tried :)"
@@504Rone from my testing it uses GPU ram only if bakeing with GPU, and CPU ram if baking light with CPU. In both cases on s large scale it is very easy to runout of memory
@Dmytro Momotov, similar to having high frequency and low frequency reflection probes you can do the same with APV baking different sets for high medium and low, similar to how the auto placement works from sky to detail, but you have a lot more control over min max distances for those specific high mid low areas.
Carefully set the lightmap resolution scale for all assets (you probably have many assets with way too high a scale driving up the memory needed to bake) and separate the large scene into multiple smaller scenes.
Unreal is spitting features after features that is breaking technologies. Unity is needlessly trying to compete with them which they will miserably fail. At least be master at what you good at. Please work on 2d features where unreal can never compete. 2d lighting and shadows, their performances and technologies related to 2d that where you should directionally focus. Also don't forget that your user base are indie. Give features that will benefit indie teams. Indie developers are somewhere forgotten in the dust it seems. Please work on engine features that already need to be polished. Am i wrong please correct me.
Unreal's Nanite and Lumen are groundbreaking, but they tend to crash the app on low-end laptops. I want Unity to make more stable versions of those techs and finalize their GPU lightmapper
Yes you are wrong because you don't understand that Unreal's features are merely a marketing bullshit. Not optimized and not fit to be usable on regular hardware in a video game.
This is NOT efficient! I rather have control of my probe placement and i will know what probes are affecting what in the scene. This feels like a replacement for an ACTUAL GI system, which you STILL need for your scenes
Maybe it needs the grid-like placement for easier per pixel evaluation (as if it were a volume texture? I don’t really know under the hood). That said, for hand placing them, I will still be using the older one.
Lighting is one of the complex subjects that a lot don't get right or just don't bother to get right due to its complexity.
It would be nice if Unity expanded the Unity Learn section on lighting.
Using the new APV system. How to create nice lit scenes from A to Z.
From smaller scenes to larger scenes. How to do proper Day / Night cycles.
All for URP and HDRP.
Exactly, we end up having to scour the web for tips and tricks from different sources and attempt to compile them to work for our scene. Usually "good enough" is the level of quality we end up with lol.
I recently started learning about lighting, and now I understand why it is a separate skill.
Efficient? Not at all, there is no adequate runtime leak prevention tech for APVs. The offline touchup volumes don't do it enough justice in a lot of cases, because there are many scenarios with valid, visible rooms on other sides of the thin enough walls. You're just throwing artefacts from one side to the other, as even visible at 14:31. Plus, for our artists it's a hassle to control these volumes and use debuggers, manipulating the grids of probes feels easier for them. And biasing in itself introduces a lot of caveats as well.
A runtime depth-based or visibility-based invalidation of probes is necessary to determine if a probe at a sampling position should be contributing. Though, understandably, that may require a lot of storage (L2 SH needs 27 floats and there are packing optimizations, and now compare that to 64 octahedral depth samples). In fact, you tried to experiment with octahedral depth and even have semi-working lightmapper output support for them, but I imagine you didn't implement it precisely because of the storage concerns, am I correct? Heck, how about just using a rotated cube depth encoding like in TLOU2? It had very sparse hand-placed probes, but APVs can have so much more density and give accurate results, so why even not? Or is it a problem because now you can't do hardware interpolation of the probes in a brick and will require 8 full loads?
But other than that, it's a solid tech overall, but runtime leaking prevention is absolutely necessary in order to make it truly useful for us and, I believe, many others.
Unity investing development time in already obsolete technology, instead of working on developing a proper realtime GI solution...
@@thomasfalcon6350this
@@thomasfalcon6350 I can assure you that APV is far from being an obsolete technology, as realtime GI solutions in themselves have loads of limitations. Something like Lumen is a cool piece of tech, but you sure as heck ain't gonna run it on a last-gen conole, mobile phone or a low-end desktop GPU. And good luck fighting the noise and temporal artefacts, especially if your project cannot support mandatory TAA. And also good luck having reduced precision and artefacts for tranparencies due to them relying on Lumen data projected onto the camera-relative APV equivalent in Unreal.
@@thomasfalcon6350 Most game/realtime 3D projects are not aimed at platforms/specs that can support realtime GI at respectable frame rates in the coming... Lets say, 5 years? Likely longer if we take mobile and standalone VR into consideration.
So from that perspective it's reasonable to suspect that APV will remain usefull for years to come if they work out the kinks and continue to support it. And besides that, HDRP already supports raytracing, specificaly path tracing, that once the hardware of the average joey catches up to it, already provides your desired proper realtime GI solution.
So your claim that this is a "obsolete technology", i do not understand how you could say that, and i find my self on the other side of that argument. I think this is a great piece of technology that fits the bill for the current requirements untill we likely reach a point years from now, where most to all hardware can finaly provide realtime GI for all users. Untill then, things like APV are required to fill the gap.
@@thomasfalcon6350 Realtime GI is too expensive for mobile platforms. Besides, the most impressive lighting in AAA games also still depend on lightmapping and baked GI. It is far from obsolete. The fact that the APV system supports per pixel lighting makes lightmapping in many cases already not a necessity anymore and help iteration time. I do agree that in some cases a better realtime GI system would be preferable, but only good GPU's can run it.
Don't be fooled by Unreal Engine tech. Things like Lumen are very impressive, but it comes with a cost.
4:29 Oh, I was expecting to be an updated tool to probe placement but the per pixel enhancement is very quite nice!
great talk thank you
long waited content Finally
So can we bake AVP and lightmaps together? Or we only have to choose one of them in one scene?
afaik you can mix them, but you have to choose either "Receive Global Illumination: Lightmaps" or "Receive Global Illumination: Light Probes" per each MeshRenderer. At least with the latest beta version that I tried :)
@@KYL3R64 Thats what I thought but thank you for sharing mate, Cheers!
16:20 👍👍👍
after baking my probes are all black. what do I do?
Lighting Scenarios is not working as advertised and it barely does anything. You cant find examples on how to blend it anywhere.
I try to do this but find out my scene always baking the lightmpat, even when i press the bake button in the probe volume menu.. that's why mine took very long time to bake, didn't find the tutorial how to bake with probe volume only. And after the bake complete, i try to add sphere mesh with hdrp/lit material, then try to moving it around to try if the APV working, but the sphere still look flat all over the room not affected with all the probes, or simply it's just my wrong step.
I think you need to enable what source of lighting will affect each object, like Kai F said above "
Kai F
2 weeks ago
afaik you can mix them, but you have to choose either "Receive Global Illumination: Lightmaps" or "Receive Global Illumination: Light Probes" per each MeshRenderer. At least with the latest beta version that I tried :)"
If lightmaps are created it means that your static objects have "receive global illumination" set as "lightmaps" instead of lightprobes.
which version of unity ?
I can't find the experimental package in the latest version of Unity. How can I test it?
I think light probes could be massively compressed. It is memory bounded, the programmers should know it. Not just memory, it's bad for build sizes.
Background noise of talking people are too loud please aware this
How to make it work with multiscene workflow??
Why is everyone talking through the presentation?
Because it is not a closed environment, it is on the show floor xD.
@@QErikNL 😓
Haha, I do find the whole situation and the comment hilarious.
At least what he is saying is clear though, enjoyed it.
64 Gb RAM is not enough for bakeing large scenes :(
This uses gpu v ram not cpu ram.
@@504Rone from my testing it uses GPU ram only if bakeing with GPU, and CPU ram if baking light with CPU. In both cases on s large scale it is very easy to runout of memory
@Dmytro Momotov, similar to having high frequency and low frequency reflection probes you can do the same with APV baking different sets for high medium and low, similar to how the auto placement works from sky to detail, but you have a lot more control over min max distances for those specific high mid low areas.
Carefully set the lightmap resolution scale for all assets (you probably have many assets with way too high a scale driving up the memory needed to bake) and separate the large scene into multiple smaller scenes.
Yeah, 👍
What version of unity does this actually work in? The probes don't display in all the versions I tried.
Isn't the amount of probes a bit excessive? Would this be inefficient for mobile platforms?
Isn't this for HDRP?
@@StevenBloomfield It's for URP as well.
@@SimonTysland URP doesn't get many of these features so as far as this presentation goes one can fairly say it's only for HDRP.
@@forasago It's for URP as well.
i don't get 60% of unity user use urp 20% built in and only 10% for hdrp but you work on it move than both of them
2023.1 tech stream just added this feature for URP
Unreal is spitting features after features that is breaking technologies. Unity is needlessly trying to compete with them which they will miserably fail. At least be master at what you good at. Please work on 2d features where unreal can never compete. 2d lighting and shadows, their performances and technologies related to 2d that where you should directionally focus. Also don't forget that your user base are indie. Give features that will benefit indie teams. Indie developers are somewhere forgotten in the dust it seems. Please work on engine features that already need to be polished. Am i wrong please correct me.
Unreal's Nanite and Lumen are groundbreaking, but they tend to crash the app on low-end laptops. I want Unity to make more stable versions of those techs and finalize their GPU lightmapper
Yes you are wrong because you don't understand that Unreal's features are merely a marketing bullshit. Not optimized and not fit to be usable on regular hardware in a video game.
lol, no
This is NOT efficient! I rather have control of my probe placement and i will know what probes are affecting what in the scene. This feels like a replacement for an ACTUAL GI system, which you STILL need for your scenes
Maybe it needs the grid-like placement for easier per pixel evaluation (as if it were a volume texture? I don’t really know under the hood).
That said, for hand placing them, I will still be using the older one.
@@alejmc Same. And it SHOULD NOT be a replacement for GI