As some of you know, there is now a fork of Godot with path tracing. Even though it’s still work in progress, it can already create stunningly beautiful lighting. (I don’t think that Gamefromscratch video does it full justice. In my own experiments it works much better.)
I hope that vendor agnostic parts of this fork become part of the core engine as soon as possible. Path tracing is not the future, it’s the present. Personally I haven’t played any games without path tracing for months, and at the moment gamers all over the world are eagerly waiting for Crimson Desert - a game that relies heavily on path tracing.
It would be appropriate to link to the fork in your original post for those that don’t know.
The problem with that is that the Godot project does not accept LLM-generated code. That alone will keeps this particular version from ever being pulled into the Godot core.
You might benefit from seeing the latest Godot Tomorrow video, where Emi talks about the fact that while the Godot team would like to support many things like this, they are not chasing AAA game building.
That’s great for you. There’s also a huge surge in retro games with simpler graphics right now. There are two things that are much more likely to determine what happens to the future of games. One, the economy, and two, the cost of hardware. The recent economic downturn has seen a drop-off in purchases of AAA titles, and an increase in the purchase of cheaper indie games. Additionally, the cost of GPU hardware and RAM has shot up so precipitously, that it is likely to reduce the number of gamers who buy bleeding edge hardware. Other issues such as US tariffs have also caused issues with the sale of new gaming hardware.
So while it’s great that people are really looking forward to Crimson Desert, the number of people who can A) afford it and B) play it on hardware for which it was intended maybe limited.
It looks quite bad in that presentation video. All supposedly path traced examples look way worse than their non path traced counterparts. Funny how the guy in the video is constantly trying to gaslight the viewer that they look better. I mean, you can call it “path tracing” all you like but if the results are worse when you enable it - what’s the point?
This is typical “ai”-induced behavior we’re sadly witnessing all around. Produce some shit using agents, and then everyone is obliged to praise it merely because it runs.
Can’t I call it path-traching? I initialized the things that it need in the code and it gives no error,sooo technically my engine got path-traching : )
Nice little demo.
IMHO, it’s only an experiment. I would like to see path traced games with full raytracing and global illumination. But it’s still too computational expensive. With a lot of frame generation, up scaling and denoiseing some impressive rendering is possible, on high end hardware. It’s too elite to use it in full AAA titles. There will be for some time no hardware available, that could handle this with mid range GPUs.
As I have an AMD, this fork is not that interesting for me. And as you can see in the reactions, that only a few people want to see AI slop in the Godot engine.
It’s ok, that a developer at Nvidia is experimenting with Godot and this tech, to find bottlenecks that could be targeted in future development. There are a lot of corner cases where this pipeline will break down (like fog). In complex scenes I hate the temporal glitch of this tech, when the lighting is calculated over many frames and the result evolves over time.
It’s not only about realistic graphics. My big hope is to have at some point a light model, that is realtime, without rendering / baking the lighting in development. It can be used in stylization, too.
But we are not there, yet. And with the current hardware development it’ll take some more time.
To me this technology would be great IF it encouraged most studios to proiritize optimization so future games can look stunning and run even on older hardware.
Unfortunately most of the industry took the exact opposite approach, to a point where as soon as I hear path tracing, I think “ahh okay, so I get to either watch a slideshow at 10 FPS, or I get to run the game at a somewhat stable 60 FPS with fake, horrible looking generated frames.”
I like global illumination with bounce lights and nice shadows, all dynamic. I hate the time of baking lightmaps or Voxel-GI. And ScreenSpace GI is not the solution. I’m coming from rendering in film and video. One benefit of realtime for me was that there was no rendering, just go. And I welcome any light tech, that can render in realtime without baking.
Do you want to gatekeep games, that are allowed to be build with godot? Only lambert direct shaded with no or only simple shadow?
Whatever we get it won’t automatically make the game look better in artistic sense. I think that’s a common misconception. Bounce light by itself doesn’t produce aesthetically pleasing visuals. And whenever a lighting feature arrives, some people always want moar, expecting the tech to solve their purely artistic problems. Imo, you’ll always need an art director to make your game look good. Good art direction can make miracles with merely lambert+ao in terms of emotional impact of the imagery, whereas the lack of artistic eyes on the project will almost inevitably make it look underwhelming regardless of the amount of rtx you throw at it. “Programmer art” comes in globally illuminated flavors as well.
What kind of question is that? Do I have the power to?
I’m not for banning AI discussions. But back to topic, with the path tracing fork. I admit, that AI is part of the solution how Nvidia will address this, but no more OT AI discussion, please. (I was part of)
Some art styles it will support. Gradients are part of many art styles. GI with bounce lights produce a lot of nice gradients instead of flat shading. You can fake or prerender a lot of it, but I would like to get it from the rendering instead of faking it. It supports the shape of the objects, like Ambient Occlusion. I like the aesthetic beside photorealism (not always my target).
A lot of my projects have dynamic objects, where the player or the simulation can place or build structures, that conflicts with prerendered light maps or Voxel-GI.
I’m not against projects in other art styles, love them, but GI has it’s own place in some styles.
I don’t think we will see production worthy path tracing any time soon anywhere. Even Unreal’s Lumen is not remotely close to path tracing. I in fact think its foolish to chase path tracing for realtime graphics until gpu architectures that are basically built for local illumination, change considerably to fully support path tracing in hardware, not as a general compute implementation on architectures we have now. It’d be a considerable paradigm shift and since nvidia is busy peddling tpus for the time being, there likely won’t be any significant moves in that direction.
