Polygons, voxels, SDFs... what will our geometry be made of in the future?

Yes and no.

Yes, because I've been following Teardown for quite a while now and it looks like a good step in the right direction. No, because the voxels in it are far too large - or far too few in numbers, as @eloyc put it.

There's also this project: https://twitter.com/ProgrammerLin/media

I understand. But I think the tech in Teardown could be extended into something that looks less "on-the face" voxelly, yet the physics and maybe even the illumnation is done on a voxelspace of similar (maybe even lower) resolution.

Like, a Teardown where it's the same world in complexity but with a graphically detailed shell over the underlying lego world driving the sim. That detail geo on top could still be poly meshes anyway, or perhaps something more experimental ala dreams or claybook, but I find the latter less likely.

I wouldn't be surprised if that ends up actually being the direction he takes his research after teardown.
 
Last edited:
The future is definitely volumetric, as Atomontage puts it.

I've lately been dreaming of fully voxelized games with physics and ray-traced lighting. Such a game could be based on the movie Tremors, for example. Those "graboids" coming from under the ground to get you...

Anyways, when will NVIDIA bring their first "voxxelerator" to the market?

Noita just came out of early access. While it's 2d, it pretty much shows the gameplay limits of a world where everything is simulated. How you'd make something like this in 3d work technically and gameplay wise is anyone's guess!

(Side note: buy Noita, it's ace)

 
Noita just came out of early access. While it's 2d, it pretty much shows the gameplay limits of a world where everything is simulated. How you'd make something like this in 3d work technically and gameplay wise is anyone's guess!

(Side note: buy Noita, it's ace)

Noita also does awesome 'store game world to HDD' trick when it is far enough from active area.
 
(Side note: buy Noita, it's ace)
I did so the nanosecond it hit Early Access back in September 2019. To date I have a couple of hundred gameplay hours under my belt robe. To me, the game is quite a bit more than just ace :)

Can you imagine Noita in 3D and in first-person? We need another young, brilliant mind like Carmack to make that revolution happen (plus a lot more computing power!).
 
Not very recent, but anyway:

I mean, look at that. On old PS4 hardware. I can only imagine the possibilities in the PS5. :p And I hope this road doesn't end up being forgotten as an exotic tech and that's all.
 
Not very recent, but anyway:

I mean, look at that. On old PS4 hardware. I can only imagine the possibilities in the PS5. :p And I hope this road doesn't end up being forgotten as an exotic tech and that's all.
well its' definitely feasible, and it showcases the power of a pure compute based engine. It would appear the only thing holding it back are the tools to scale up to develop massive sprawling games that we have today.
I think if they can figure that part out, it may well be a reality.
 
It's a shame we've not seen another major alternative geometry solution outside of Dreams and UE5. The latter will boast a fair few games I guess. Dreams is it's own little world, even if Dream 2 is, er, enables bigger worlds.

Dragon Age, Square, Capcom, Insomniac, Gorilla showings all look like current gen Ultra settings + SSD streaming + a bit of RT. Which is fine, but seems a shame that to date, Epic + those using UE5 are the only big developers jumping on something new rendering wise

Obviously early days!
 
It's a shame we've not seen another major alternative geometry solution outside of Dreams and UE5. The latter will boast a fair few games I guess. Dreams is it's own little world, even if Dream 2 is, er, enables bigger worlds.

Dragon Age, Square, Capcom, Insomniac, Gorilla showings all look like current gen Ultra settings + SSD streaming + a bit of RT. Which is fine, but seems a shame that to date, Epic + those using UE5 are the only big developers jumping on something new rendering wise

Obviously early days!
UE5 is still using triangles, it's just using compute to solve mostly everything for their geometry.
Claybook also uses SDF, I think Sebbbi could be working on something on his own time, I'm unsure of what he's working on at Unity now. But IIRC, he was seriously impressed by the artists working in dreams. And I think IIRC, he has passion to build some tools to do something similar.
 
well its' definitely feasible, and it showcases the power of a pure compute based engine. It would appear the only thing holding it back are the tools to scale up to develop massive sprawling games that we have today.
I think if they can figure that part out, it may well be a reality.
If someone could find a way to improve such tech to make possible one-piece rigged characters (instead of old-school characters made up of different disconnected parts for each limb, which makes almost impossible to achieve a non-cartoony look) and better reflections, we'd have it all. If the tech is there and someone shows that's possible, other people will jump in. Heck, even Dreams, with its limitations, is getting an important amount of attention among game developers, artists and musicians, and some creations almost look and play professional, even though it is very niche and relatively new.

Look at this (and ignore the bits you may find cheesy, lol):

Look at the pumpkin! It just looks CGI (again, ignore other things) and it's full geometry, no normal maps. Try converting that to polygons.
 
UE5 is still using triangles, it's just using compute to solve mostly everything for their geometry.
Claybook also uses SDF, I think Sebbbi could be working on something on his own time, I'm unsure of what he's working on at Unity now. But IIRC, he was seriously impressed by the artists working in dreams. And I think IIRC, he has passion to build some tools to do something similar.

Yeah, I know UE's using triangles but the way it gets those triangles is definitely alternative. I kind of dismiss Claybook in the context of future stuff, with absolutely no disrespect to Sebbi, as it's a little indie title. It'd totally forgotten he was over Unity now. Very interesting to see what comes of that.
 
Yeah, I know UE's using triangles but the way it gets those triangles is definitely alternative
Indeed, well it's certainly going to be the future of triangle generation. Everyone should be moving their way into that direction I think over time. I firmly believe this is the generation that will tip the scales in the other direction. So compute will be making up the majority of the whole render pipeline, and FF supporting it. As opposed to FF making up the majority of the pipeline and compute supporting it.
 
Not very recent, but anyway:

I mean, look at that. On old PS4 hardware. I can only imagine the possibilities in the PS5. :p And I hope this road doesn't end up being forgotten as an exotic tech and that's all.



Ek9CYlgXYAEfIvc


Ek9CkhdWkAA3eN0


Ek9DLI9XUAMjasA


Just insane
 
Geometry is king! The issue is who will design all these details in game? I dont know how much time was needed to design only 1 scene in Dreams in the pictures above.

I think with even simple light effects but complex geometry the game looks much better than with complex light effects and simple geometry. (Just my meaning)
 
Geometry is king! The issue is who will design all these details in game? I dont know how much time was needed to design only 1 scene in Dreams in the pictures above.
Who will? Whoever who wants to (or gets paid for it). At any rate, understand that most of these videos are made by just one person VS a whole team.

I think with even simple light effects but complex geometry the game looks much better than with complex light effects and simple geometry. (Just my meaning)
I don't agree. Of course, detailed geometry is essential in making a game look good, but good lighting is what makes even just a cube look realistic, though.
 
New clip from Lin.

This is the final installment of the mini videos teasing some of the features I've been working on, this time including object fracturing, a WIP voxel character animation system, new path tracing (still working on the denoiser), day/night, some new terrain generation that gives me more control, and improved scene composition.

Next video I'll be giving a long showcase of everything in the engine, showcase the water physics, and talk about the game this is turning into.
 
Yeah, I watched that on Twitter yesterday. The thing I like is that these are "true" voxels within a grid, not rotating cubes. :mrgreen:
 
I don't agree. Of course, detailed geometry is essential in making a game look good, but good lighting is what makes even just a cube look realistic, though.

Light is only second. If you don’t have a geometry which is not giving you nice informations, the games look ugly even with good light shaders. It’s the geometry which give lights and shadows the unique touch. So complex geometry with simple light will always better than simple geometry with complex light.
 
Light is only second. If you don’t have a geometry which is not giving you nice informations, the games look ugly even with good light shaders. It’s the geometry which give lights and shadows the unique touch. So complex geometry with simple light will always better than simple geometry with complex light.
I'm sorry, but that's a totally subjective opinion with which I don't agree at all. Of course, detailed geometry helps in achieving a realistic look, BUT without realistic lighting you go nowhere.

I already used a simple example: a single cube with realistic lighting may look real enough. On the contrary, put a very elaborate sculpture, for instance, in a room lit by a standard point/spot light with no GI, ugly shadows, etc... It just doesn't look right, it looks gamey.

Some modern games feature very high poly models but even then, if lighting/shadows are not good enough...
 
Back
Top