John Carmack and QuakeCon

Started by rcallicotte, August 06, 2007, 10:38:59 pm

Previous topic - Next topic

rcallicotte

Have you guys seen the map editor for id's next generation game engine?  OMG.
So this is Disney World.  Can we live here?

old_blaggard

http://www.terragen.org - A great Terragen resource with models, contests, galleries, and forums.

Will

The world is round... so you have to use spherical projection.

rcallicotte

Apologies.  It's a huge download at FilePlanet - http://www.fileplanet.com/179536/170000/fileinfo/QuakeCon-2007---John-Carmack-Footage-(HD) or an easier one at Fileshack - http://fileshack.com/file.x/10928/HD+Rage+FileShack+QuakeCon+2007+Footage

I know this sort of thing has been played down, but what we're learning now with Planetside could be the new game interface of the not too distant future. 
So this is Disney World.  Can we live here?

Oshyan

An interesting and worthwhile viewing, but pretty much the entire sales pitche focuses on "megatexturing" and the tools to take advantage of that. Yet the results shown, while nice-looking, were not particularly outstanding in my opinion. Too much is still done with textures, things look very flat still, and even with the megatexturing technology the sheer time-investment necessary to truly capitalize on it and make every pixel unique is just too daunting in a practical sense.

Carmack seems down on procedural approaches but they're really the only way to get that level of uniqueness on the scale he's talking about. You can have fantastic artists plugging away at a 64kx64k texture for months and while the end result may be better overall than something done procedurally in a day or two of calculation, if you combine the two in a better balance you would really have something more special I think. Of course for the procedural approaches you spend a lot of time beforehand in setting up the systems to do it, but once they exist you can re-use it as much as you want and get unique results every time. That's the power of Terragen's texturing approach and is a big reason why it tends to look better and more realistic than some other systems that rely more on texture tiling. What's ironic with this megatexture thing then is that despite the ability to have each pixel be unique, the artists are still working with "decals", "stamps", tiled textures and other repetition, so fundamentally there is little practical difference.

In any case I've always been interested in what Carmack is up to and often impressed by what he is doing, but I think in the last generation or two he has fallen behind others to a greater and greater degree as far as the actual practical results of what he's doing. His tech may still be the flashiest under the hood, but these days Monolith (who made F.E.A.R.), Epic (Unreal Engine 3), and Crytek (Crysis) are all at least in the same neighborhood as Id and by many estimations a bit further.

I for one have been a *great* deal more impressed by what Crytek has done with Crysis, both on the game side *and* - most especially relevant to your points here re: convergence - on the tools side: the editor functionality. I'm fairly sure you've seen the Crysis editor videos that are out there (link in a moment). While Id may have a great pipeline for making big, unique textures, I didn't see anything that allowed you to edit terrain or even models in-editor. I didn't see anything about adding massive amounts of vegetation. I didn't even see anything to do with dynamic shaders (water for example), though I'm sure that stuff at least *is* possible through their editor.

But what Crytek can do is so far beyond. You can edit the terrain in a fully volumetric way in the editor, creating caves and any other feature. You can use pre-created modular models to create unique buildings in realtime, without having to use an external model editor and lay down any other object you want just as easily. You can use population-like tools to spread out vegetation of all different kinds quickly and easily. Perhaps most impressively you can lay down roads that conform to the terrain and then actually modify the terrain to accomodate the road dynamically. To me Crysis is head and shoulders above what is shown in these Rage videos, and if anything is showing us the possible future of tools like TG2 it is the Crysis game and engine. Take a look:
http://www.gametrailers.com/game/2509.html

Pay particular attention to the editor demos here:
http://www.gametrailers.com/player/17743.html
http://www.gametrailers.com/player/17762.html

On top of everything else Id is doing, and on top of the power of their editor, in terms of pure engine features they also have fully dynamic GI-like lighting, atmospheric rays, destructable environments, highly accurate colission detection and model deformation (plants move when you walk through them), etc, etc. Some of those things are probably in Rage as well, but Crysis brings it all together. To top it all of it's coming out this November, and Rage, as I understand it, is still a ways off. I think Crysis is as close to a "game changing" product as we've had in the games industry for quite a while. If I were going to license any engine to create a game that would definitely be it.

- Oshyan

ProjectX

If I were to back an engine to be picked up by the modders (and hence be a good indication of tool-power and friendliness) I'd probably back Unreal Engine 3, not just because there are tons of games using it and Epic has a great history of mod-support and being mod-friendly, but also because it really is quite easy to use, and while crytek has a lot of good tools (at least this time around) they only provide assets for jungle environments really, and all the realistic modder (the ones that actually release mods) usually start by using existing assets and then gradually replacing them with their own. Id tech 5 might get something, since it looks quite easy to use, and they claim they learned their mistake from Doom 3 (making great tech, but no real tools to use them) and the mega-textures are still light-years ahead of everybody else in detail, even if they do use texture swatches, but it's too early to tell really.

As for professional game engine dominance, I think Epic has had it in the bag for quite a while now, and if they survive the lawsuit, then they'll have it in the bag for generations to come. The last major engine licence ID got (AFAIK) was Call of Duty 1, and that was with a modified Quake 3 engine, crytek have never had an engine licenced, but Epic have had consistent licencing from each generation. To be honest though, a company's history will do more than their technical excellence to sell engines.

From reading the features and looking at videos, Crytek's engine is the most visually stunning, but until I go hands on with their tools, I'm not believing a word they say about ease of use or flexibility. ID tech 5 is next-gen (as in the next generation, not this next-gen) and so doesn't really fit with the other engines, but if they were to compete I'd say their engine boasts the most power and a lot of ease of use, but as has been said, we haven't seen much of it yet, and Unreal Engine 3 has the greatest flexibility and enough power for this generation, but not enough for ID tech 5's generation.

I think it's going to take maybe 2 or 3 iterations (generations) of games before procedural becomes quite mainstream, and I think in the long run it'll be the engine that encompasses artistically editable procedural generation (generate it procedurally, then modify it as if you were painting) that wins in the game world.

That has got to be the most confused and rambling post I've ever made. Sorry, no maps for you!

normhol

Thanks Oshyan. Those are very enlightening trailers. It's obvious I was born 60 years too soon  LOL
Norm. ::)

rcallicotte

ProjectX, that was interesting.  I believe you're right about what will do the best in the future - procedural graphics that depend upon an artist's ability to create and not a technician's brain to calculate.  Meats said this well, when he explained about zBrush's usefulness.  Programs that are technically fantastic and allow the artist to simply create are what will do the best, logically.

As far as who will be on top, this is an interesting challenge to discuss.  I have opinions only, but my guess is something will come along and change everything even more than what Crysis appears to be doing now. 

I agree with Oshyan that Carmack's engine is cool, but not the next best thing.  Nevertheless, I would imagine it will do well in the marketplace.
So this is Disney World.  Can we live here?

rcallicotte

Oshyan, thanks for your point of view.  I've seen this before and actually have pre-purchased the game, something I've rarely done.

Has Crysis used procedural graphics on any stage of this editor or the game itself?   

I agree about procedural graphics.  After half-stumbling through a very complex book you recommended weeks ago, I have been able to surmise that procedural graphics is still in its infancy as far as the general public is concerned.  Perhaps there are geniuses in golden towers somewhere working on virtual worlds only secure enough for MI6.  LOL   ::)  Imagination at work.   ;D
So this is Disney World.  Can we live here?

moodflow

Yes, procedural is the way to go for sure.  As Oshyan stated, there is more work "up front", but once thats done, the results are way more impressive, flexible, and variable.
http://www.moodflow.com
mood-inspiring images and music

ProjectX

The problem though is losing full artistic control over things, and also the problem of not really having anything eye-catchingly man-made. A lot of things can be made procedurally, but we will always need designers for the final step. That's why a mix would be the best option, you create it procedurally, save it, then edit it manually.

rcallicotte

Maybe this is because where we are right now.  The potential for doing it all in one operation procedurally is looming its inviting head.  TG2 is a good example of what "Texturing and Modeling: A Procedural Approach" says is an advantage of procedural textures over an image - "A procedural texture can be parameterized, so it can generate a class of relaxed textures rather than being limited to one fixed texture image."  Procedural textures can also be more compact than their image counterpart, which will change the face of what can be done (with less).  Didn't someone write a 75K program recently that was procedural in nature and played better than the first Wolfenstein?



Quote from: ProjectX on August 10, 2007, 03:10:17 pm
The problem though is losing full artistic control over things, and also the problem of not really having anything eye-catchingly man-made. A lot of things can be made procedurally, but we will always need designers for the final step. That's why a mix would be the best option, you create it procedurally, save it, then edit it manually.
So this is Disney World.  Can we live here?

Oshyan

August 10, 2007, 05:38:48 pm #12 Last Edit: August 10, 2007, 05:47:11 pm by Oshyan
ProjectX, I may not be as familiar with the various engines in question as you (I used to follow this stuff a lot more closely), but I'm curious from where you draw some of your information. For example I have only seen a few bits of info on Id Tech 5 (Rage game engine), and from what I've seen it's *not* really more impressive than either UE3 or Cryengine 2, and that includes the "mega texturing". You get up close to the textures and they still look crap. Things like skeletons are still done with flat textures. Tiling and repeated textures are still heavily relied upon because of the sheer artist effort involved in creating 64kx64k unique textures (let alone larger). This is why improving procedurals is an important component (note: I do not think we should rely exclusively on procedurals).

Meanwhile both UE3 and Cryengine 2 have soft shadows, high polygon counts, normal mapping, HDR lighting, DoF, motion blur, etc, etc. Cryengine 2 even has realtime GI-like tech, amongst other seemingly unique technologies. I still maintain that Cryengine 2 is one of the more impressive offerings, and I don't think there is necessarily much to suggest its possibilities will be limited to jungle environments, except of course that the default assets will be tuned toward this. But is this any different than saying the default D3 or Source assets were tuned for their certainly also limited environments (particularly D3)?

So with that in mind I wonder on what information you are basing statements like: " I'd say their engine boasts the most power and a lot of ease of use, but as has been said, we haven't seen much of it yet, and Unreal Engine 3 has the greatest flexibility and enough power for this generation, but not enough for ID tech 5's generation."

Also of note is that the Doom 3 engine *was* licensed for several games, most notably Prey:
http://en.wikipedia.org/wiki/Doom_3_engine

That being said certainly UE3 has attracted far more licensees:
http://en.wikipedia.org/wiki/Unreal_Engine_3#Unreal_Engine_3_2

But of course we know that this is somewhat by design - Id has by their own admission not focused on tools in the past, while Epic most definitely has.

Cryengine 2 has also had several licenses purchased recently, which is notable considering they did not license Cryengine 1 and Crysis isn't even out yet. I think that speaks well for their system, and we may expect to see many more licenses to come:
http://play.tm/wire/click/1457709
http://play.tm/wire/click/1420713
http://play.tm/wire/click/1334542

So basically what I'm saying is I still think Cryengine 2 is the most impressive engine all around that is currently available, and that includes editing possibilities. UE3 hasn't shown anything close to the amazing ease of use and power of the Crysis editor, at least when it comes to outdoor environments (which are a lot more interesting to me, especially considering we're here on the Terragen forums discussing it). Rage is maybe too early to tell, but I remain not particularly impressed by "megatexture", and though I think it's a natural evolution, I don't think its value is necessarily in vastly enhanced detail or uniqueness for textures. The massive bottlenecks in the content creation pipeline still need to be solved and that will require heavy investment in realistic, directed procedurals.

- Oshyan

ProjectX

You make many a few fair points. I noticed all the effects you described (apart from GI) in the Rage demo, but they are quite hard to see, but they are definitely there. The skeletons you saw in the editor were textures, yes, but these dase (especially with displacement mapping) textures are a lot less 2D than they used to be. You could argue that the entire terrain generation system of Terragen is one big Displacement mapped texture. High polygon counts were certainly noticeable in the animated garageworker (I think he works in a garage) in the rage demo. They didn't mention the effects, but they are there.

" I'd say their engine boasts the most power and a lot of ease of use, but as has been said, we haven't seen much of it yet, and Unreal Engine 3 has the greatest flexibility and enough power for this generation, but not enough for ID tech 5's generation."

I say this because there were tons of effects shown in Rage, as well as Megatexturing, and it really is in a generation after the current engines. Like the Doom3 engine however, it will probably be the worst of the next gen engines, simply because it came out earlier. I'm starting to think John Carmack is working in his own slightly shifted timescale of generations compared to everyone else.

I've used the UE3 engine, and I must admit that i can only really base my experiences with Crytek with CryEngine 1, and the videos that I've seen, and while Crytek has always won hands down on exteriors, their interior editing was rather restrictive, whereas UE3 can manage both relatively easily.

I like to think of it as a scale, great exteriors on the left, great interiors on the right. Crysis is on the far left, ue3 is in the middle and the Doom3 engine would be a bit further to the right (I can't compare ID Tech 5, because I haven't seen them build an interior, or exterior - but as a guess I'd put it somewhere near the middle). I'd bet my money on the middle of the road engine doing better than the one that handles one aspect really well, but doesn't do so good on the other.

Of course when Crysis comes out I'd love to be proved wrong, and be faced with a wonderful and easy-to-use editor, but I can't see it happening yet. I don't think you'd find a 2D side-scrolling mod for Crysis.

As for licensees, I must admit to being a bit behind on who has licenced what, I'd completely forgotten about Prey being Doom3 and hadn't even noticed Crysis getting any licences, so for that I was entirely wrong.

I still stand by Unreal Engine 3 for being easiest to use, if only because I have used it and found it easy to use (their Kismet system - a graphical, node-based, method of creating scripted sequences without the need of a programmer - is just beautiful, especially for a terragener like myself). All we've really seen from Cryengine 2 is that it can make some very nice terrain, we haven't seen material editors, scripting editors, particle editors, etc. All of which you can not only see, but you can use for 15 bucks and a copy of steam.

btw. As for what calico said, yes I totally agree there is a lot of potential for procedural textures (although my Tb harddrive doesn't give a **** about texture sizes) and even procedural modelling. Roboblitz, an Unreal Engine 3 XBox Live Arcade game used procedural textures in order to reduce their filesize from around 800 meg to 50 meg, but even they had to use some artist-made textures, because not everything can be procedurally made, or at least, not easily procedurally made. Actually, I've used the software they used, it's the Mapzone 2 editor (which i mentioned on the file sharing area) and it is incredibly powerful, but not as powerful as a good artist with a wacom and photoshop. That's why I stand by my statement that when procedural generation is used to generate content that the artist can then edit, that is the ideal engine.

rcallicotte

"That's why I stand by my statement that when procedural generation is used to generate content that the artist can then edit, that is the ideal engine." - ProjectX

Agreed.   ;D  And I'll add "...that the artist can edit intuitively and readily..."
So this is Disney World.  Can we live here?