It's definitely possible. Firstly I've described my own solutions (varied displacement of one larger texture instead of repeating the same texture, for example). Secondly, and more importantly, there are lots of examples showing the progress of procedural content. There's Uncharted with its animation blending, allowing for hundreds of animations that give Nate a fairly natural look without needing the artists to create hundreds of individual animation cycles. There are the procedural dungeons of Champions of Norrath that show how predesigned tiles can be assembled into gorgeous random levels. There are the random worlds created in Civilization Revolution and Captive and Elite that add considerable replayability of the games. There's SpeedTree building varied forests from parts instead of duplicating the same few tree models ad infinitum. There are the procedural creatures of Spore, that animate without anyone having to apply human intelligence to how the creatures should move, instead creating the animations on the fly from a set of rules. What reason have you to doubt that computers can select and combine various pieces, or apply shaders to vary looks?
I am not sure I would call procedural art to Uncharted animation system
It seemed to me it is more some kind of interpolation, it is like calling Anti Aliasing procedural art.
Champions of Norrath, Civilization Revolution and Captive, if I am not mistaken are pretty simple in gfx and in terms of environment, so also a very easy to create a rule set and randomize stuff, completely different matter in programming terms.
In other game like RTS maps actually it is a nice addicition, sure it is much less pleasurable to play ina computer generated map, but it is cool to play sometimes.
But they are just that, a nice addition not a replacement.
On Spore I kinda agree with you, but lets remind that it is a game 8 years in development and the use is limited to animation and unrealistic creatures. Personaly I wouldnt mind having a look at the basis fur such generation but for the little I saw it seems that is quite physics based, not "art based". Anyway a phenomenal work that probably couldnt be used in many other games, specially more realistic ones.
This is the kind of stuff that we should see more times.
But you have conscience that it is creature made by you, so like in a blog you get "good results" from the pre-made stuff but you don't get that distinguish look that make that a worthwhile job to be shared (unless the text is good).
None will remember spore because of the animation, but they will actually see it as tools for simpler art works.
See the above. There's clearly different levels of procedural effort. Some things would need a load of complex AI effort draining resources (like From Dust's procedurally created worlds). Others can be jolly simple, like shifting an NPC's base hue or value so it's a slightly different skin-shader to a copy of the same model, just to add a little variety.
From what I saw of From Dust I wouldnt call that prucedural art, it actually seems symplified tools, the same kind of tools there was in the Far Cry XB(1) map generator.
Maybe what you call procedural art I call of tool.
Perhaps in some cases. But even if so, isn't that better than seeing the same 10 people over and over?! Why is the argument against procedural content "it's not perfect", when we accept so many other tricks and imperfect solutions in our realtime engines?
More tiresome than seeing the same models over and over?
I dont mind reading/hearing/seeing/... a
good book/movie/album/... several times. But something that is only passable it gets really tiresome very fast. Tell me, you usually see/hear/read again stuff that it isnt good (only enough)
You're still focusing on human characters, and I already said that'd be the hardest nut to crack. Are we to abandon all progress in procedural content creation because we can't nail human's down first try??
I think it would aply to many thing not just humans, but to at least to everything that we have a depper connection, or that we can easily see patterns or single parts (like goblins legs).
Anyway I do think R&D should be done, just think that tools would be at least as important and in the near terms, much more important to actual games
See SpeedTree. You store a dozen small pieces in RAM and these are assembled on the fly. That's a lot more RAM saved versus storing lots of individual models. Or see Uncharted - you only need as much RAM as each of the individual motion path, and not enough RAM for every combination of every animation blend.
That is where it is perfect IMO ( I think I said so some posts ago).
But mostly because it stress to much or perception, these are offline rendered, complex objects that come in large bulks and is really hard to see patterns in it. Also we dont care about details in trees, like we do in human made stuff, so it gets even harder.
Great results too.
Are you calling me overly optimistic?
I have every faith in software developers, if only they'd be given the room to maneouvre and motivation.
Me too. But I also have in good tools and good artists
.
But that doesn't solve many of the issues. Also most of those tools you love are automations using computer AI to do work.
eg. Photoshop's smart fill is using AI to do the work of an artist.
I wouldnt call AI to that either.
Contextual fill is the exact opposite of what you want, it doesnt create any variety and dont create anything at all at the best it create continuity.
Second, when you use it you have a really good idea of how it will end.
Third, it is human controled, case to case, that is why it is powerful.
I will argue aesthetics (my, not finished, master thesis theme) and say that what it creat is not art, but the choice of use it to create/replace is art.
Except you need artists to create 500 rooms with a variety of furniture, furnishings, and paint-jobs, and storage for those rooms. Whereas using an idealised procedural system, the artists would only need create a set of components that could fit in a room, and the system would generate the rooms on the fly. Less storage, less cost, even if the experience for the user is the same. Where this wouldn't work (presently) is something like lightmaps, so you'd be limited to a simple lighting model say. Next gen's GI approximations should solve that problem.
Are you sure it is that diferent from a artist work, they reuse lots of stuff across different places.
You're looking at that from the wrong POV. The better examples are fantasy games where you have the likes of "Sharp Sword of Smiting +3". Some games advertise 'thousands of weapons and items', which present the player a balancing act in choosing which gear to use, and gives them the ever-present excitement of finding more and better loot, but it would be totally unfeasible to have artists individually create and name 10000 different items. By using a procedural system of prefixes and suffixes and base terms, the names remain descriptive and the items varied. Now of course, the items aren't terribly different - just a variation in numbers - but that difference is the basis of any loot game. A gun that does 10 damage is a great find early on. Then you find one that does 22 damage and that's a great find. Then one that does 50 damage. Each new weapon adding to your power contributes to part of the whole experience, and this is achieved without having to invent 10000 different unique items. If the content of these games was only artist created weapons, there'd be a handful of items and you'd loose the whole loot-finding experience.
I will not comment because I dont know the specif game, but t seems quite simplistic to me from your description, not sure it does have many advantages.
Given the costs of producing and rendering many different characters or level contents, the advantages of turning some processing power over to the task are quite apparent. It's a matter of how much and how applicable the processing power there is, and whether the software techniques are developed. This field hasn't seen as much work as others such as lighting solutions and AI.
The same can be said for tools.
Anyway I think procedural art is still very limited but have lots of potential at least in specific tasks like tree or spore.
I still like a sentient touch in games so it would never replace it for me, I will play much more a creatively, intelligent and unique map (like in a Mario game) than a a random, rule based one in those games you mentioned.
So they should invest in it as it can bring some great results in very specif tasks with little art in it.
