Unreal Engine 5, [UE5 Developer Availability 2022-04-05]

Maybe Tekken 9 😂

Tekken 8s visuals likely don't have anything to do with the hardware or engine, but the resources of the developer and staff and artistic intention

It's possible it would have been too cost prohibitive to dev resources to create every character and area with the same amount of fidelity and so scaled back their ambitions a bit on the whole after creating that "proof of concept" trailer.

Of course that's just one of my speculations. There is clear difference between the first trailer they showed and the latest one. Not just in detail and effects but outright art style, to the point where it seems like something they could do if they wanted, but thought better of when they actually went to make the real game which also has to run on a variety of different hardware configuration besides the hardware of PS5 and above. In addition to the dev resources argument.
 
Last edited:
to the contrary, if devs could use nanite to create characters too, they would not have to care much about LoD optimization, since it's handled by the engine.
 
Nothing about that video says "nanite on skinned meshes breakthrough that got sent to this random guy only and nobody else". Looks hacked together, maybe using a mesh sequence.
 
to the contrary, if devs could use nanite to create characters too, they would not have to care much about LoD optimization, since it's handled by the engine.

Do fighting games even implement LOD for their characters? Hell, do they do it even for the environments?
 
with nanite it's the perfect occasion to use LoDs since it's done by the engine itself ! To make the characters as close to the original CGI models as possible !
In Tekken 8 you can still notice polygonal edges on the shoulder of Jin and Law for exemple.
 
As for nanite skinned characters in general, its bound to happen at some point. The first slides explaining nanite technology stated under future work: Vegetation, skinned meshes, deformables, terrain, transparencies... "Nanite everything". Epic explained they tackled static geometry first as that was the easiest problem to solve.

Nanite is a mix of multiple technologies:

> Segmenting large models into smaller cache-friendly clusters.

> Performing occlusion culling and rendering these clusters with GPU compute using very few draw calls.

(this first two are already how AAA engines have been doing their geometry on PS4/XBOne games, UE5 only played catch up on that one)

> Rasterize geometry into an ID buffer first, before generating a full G-buffer from it in compute (this is what helps make their software rasterization be faser than HW)

> Generate a lower LoD for each cluster, and then re-segment the full mesh into new clusters, repeatedly until there are dozens of different LODs. (Thats arguably the most innovative part of Nanite)!

(Up to here, none of this is necessarely incompatible with skinned meshes)

> At render time, select the proper LOD for each segment of the model based on average poly size on screen. (this is what becomes the hardest once you have mesh defformation)

Also, UE5 adopted custom Geometry encoding formats that achieve better compression ratios and perhaps may even perform better in their compute passes (less data = less bandwith) And supporting skinning weights may also require a different encoding scheme for animated models.

In this context, half-assing nanite for characters is not impossible at all. Just a matter of time. And there would already be large preformance benefits to implementing just the compute-based cluster culling and all that. Just that would be a great start. It may not have ALL features of static nanite but still may useful ones.

To get the LoD magic working would take more work, and smart sollutions, but its also not impossible.

UE5 is aming to nanitetify everything it can, and I think we'll be there sooner rather than later.
 
Last edited:
To get the LoD magic working would take more work, and smart sollutions, but its also not impossible.

If they can accomplish this without geometric "swimming" that'll be a massive accomplishment. Every time I hear about someone attempting something like that with skinned meshes, I think way WAY back to Shiny Entertainment's game "Messiah". That implementation fell far short of their promises of highly variable and constantly changing LOD detail for models in game via their tesselation technology adding and removing triangles when needed. Its shipping implementation ended up showcasing enemy models with what I call "swimming" geometry.

Regards,
SB
 
If they can accomplish this without geometric "swimming" that'll be a massive accomplishment. Every time I hear about someone attempting something like that with skinned meshes, I think way WAY back to Shiny Entertainment's game "Messiah". That implementation fell far short of their promises of highly variable and constantly changing LOD detail for models in game via their tesselation technology adding and removing triangles when needed. Its shipping implementation ended up showcasing enemy models with what I call "swimming" geometry.

Regards,
SB

And that's kind of what the preview looked like too. Plus some straight up glitched up polys.

I consider it a miracle that static nanite LoD switches don't look more jarring than they do as they are right now already, considering they have zero transition trickery (like fading in/out)

I wonder if epic will ever try to implement real time Catmull Clark sub-d. Its the industry standard for animated charactera in offline animation... Seems like a way more elegant solution.
 
Yeah nothing impressive here, at least if there was no transition during tje flight to the moon it would have been more interesting, just like this almost one man made old game

 
"move and look Just like real skin" :LOL:
And "will all games look this good in the future?" "The answer is....NO, this technologies have been developed exclusively in Shiny Entertainment" :LOL:

I love commercials of the 90s.

The tech behind this game was truly something else for its time though. It implemented technologies that many games had years after. Too bad the game didnt receive enough success and praise and thus all the effort put into the engine vanished. It was very underrated.
It was very fun and originak. I wonder if the reason it was cancelled for PS1 was because some features such as tessellation and shadows couldnt be implement by the hardware. I cant imagine a PS1 game having these kind of advanced shadows and "infinite" viewing distances.
I wish someday we will see footage of how far it went on the PS1. They were really promising revolutionary visuals for such an old hardware and paradigm shifting. I was looking forward to see it there. I ended playing it on PC, which contributed to my curiosity
 
I'd say footage in that video was from ps1 version, the texture warping effect was an exclusive ps1 "feature" at the time.
Where did you see texture warping? The footage is so bad that the interpolation and artifacts from the video tape could be anything
 
Back
Top