Jak & Daxter 2 - high quality screens

Yeah. The cool thing about LOD in J&D as well as in R&C is not the LOD per se, but how it works. Procedurally. It's quite impressive to see 3D objects gently expand and gain more detail while you get closer.

 

True true.

 

The LOD in R&C isn't that good.

J&D(LOD)>>>R&C(LOD)

 

Folken, what's wrong with the LOD in R&C? Upon playing both games, I didn't really notice much different, other then that they both deliver an insane drawdistance, with no pop-ups what so ever...? :?

 

Weird that they worsened the LOD in R&C, it builds on top of the J&D engine ... there might be an explanation though...

 

It's a whole lot easier to LOD more or less generic landscape/terrain as opposed to a say - skyscraper filled city. (actually I don't think there has been anything even resembling a general approach invented for the latter as of yet).

At any rate, J&D's lod system is remarkably simple - much more so then I thought, and seems to me like it needs a lot of artists hand-input to get good looking results.
Using the engine alone wouldn't necesserily mean LOD system is even being used (or used well).

 

10-15 k models are for the RT cutscenes
and for phil no bump mapping in jak2

 

and you are..........

anyway ... as long as they keep progressive scan i'll be happy... at the end of the day at the size the characters are most of the time, those 15000 polygons (or less, whatever) will not really be seen so..... at least with pro-scan u get to see more of it, it's more like a shortcut to make the game look really good...

 

Is this confirmed in any way? Playing part 1, I would think that the characters in cutscenes are no different than the ones being used in-game...
Anyway, it should be interesting to compare these figures to the ones being used in the first part.

 

no bump confirmed on a french forum (by Didier Malenfant ,coder at ND )
and character are 10-15k in cutscene only.(Bump maybe in their next game)

 

Weren't the last games ingame models 6k polys? or something like that.

If so I'd say the 10K figure must be for high detail ingame(since it looks better than the previous one.), and the 15K must be for cutscenes... or maybe not...

 

Marc :

nothing technical from this forum (yet).

OT :

a job proposal a ND seams to hint that maybe they are preparing something relative to a NextGen demo (?) :

 

Actually I know quite a few who'd fit that role (yours truly included)... They're alot more common than you realize. However the due to the rather hectic schedule of production these days and the rapidly increasing demandso on art assets newer games are placing on studios, many can often be a bit lax on their hiring requirements in order to get people in, filling positions, and cranking out content.

Those requirements by the way, aren't really that stringent when you consider that students coming out of a decent CG program at an art institute or university pretty much fill most of those requirements (with the possible exception of the experience part)...

 

more like ps2...

I dont think that's next gen... Modelling 50-100k faces in Maya before reducing to 10-15k for real time via some fixed subdivision surface scheme seems pretty achievable - especially for cutting edge tech from ND programmers..

I was looking at the NV Dawn demos the other day - especially the wireframe versions, and it struck me that it would be possible to implement them on the PS2, by relying on the fact that due to the incredibly fine mesh, vertex shaded simple polys would probally suffice for all apart from the real close facial zooms...
( Before everyone jumps in about picture quality - this is just an observation about how things seem to be approaching renderman styles, with individual polys almost at the single pixel sizes already... )

 

But that's going exact opposite way of how it's Normally done...???
From what I know about modelling with various subdivision tools, the artists work on lower poly model version majority of the time, only checking the results in full subdivision/high poly (Not unlike modelling with various high order surfaces actually).
I don't exactly recall even highend CG production artists to hand-model on that level of detail, let alone games of current gen.

 

old fashioned way...

If you look at the normal mapping methods, then the hi-poly modelling followed by geometry reduction is also a valid method. ( and, as a programmer I tend to sometimes have a baseline expectation in terms of artists and polycounts.... )
The use of subdiv surfaces in modelling packages is sometimes done in your manner more to allow realtime edit & preview on slow workstations rather than due to any underlying drive for efficiency. Also I'd guess that any implementation of basic HOS on PS2 would be unlikely to map exactly to the Maya/3DSmax methods in use..