How far can graphics go?

[Flammable]

all you have to do is look at flightsim over time....last version (fs2004) looks better than FSX if you have the correct add-ons

 

[trelin]

Disclaimer: I'm not complaining here. Also, I'm not an animator, texture artist, modeler, or whatever. I'm just a random joe that notices too many details. There's also a lot I've never seen, I don't claim comprehensive knowledge, this is just off the top of my head.


Textures: Seems easy enough but my biggest pet peeve with new games are poor quality textures. Take a game like Max Payne 2. Very low poly stuff, but the textures were so clean/crisp it never bothered me. I remember one scene in particular I got killed just admiring the quality of a large billboard inside a hotel/office. Keep the textures clear and work up all the other effects from there (I'd rather have lower detail textures than high detail/high compression artifacts). A lot of this goes down to good art direction dictating how well all the textures work together. For example, Dreamfall didn't have particularly high detail textures, but they worked well together.

Biological/organic animations: my biggest wish is for randomized animations, or at least more versions of each animation. It gets tiring seeing exactly the same skeletal changes every time a character jumps/gets shot/whatever. They're just so... stiff, like gears with strict articulation, and no sense of weight. A good comparison is to synthesized music: if the beat is absolutely dead-on every time, it gets fatiguing; reduce the "accuracy" of the synthesized beat and it begins to sound more natural. A good example of stiff animations is Halo (1,2,3); when a person jumps/falls it always looks exactly the same, you can literally see where the animation sequence starts and stops. Also, when a standing person looks round, their feet don't move, like they're floating. This is an area which TF2 has done quite well, all the characters have pleasantly fluid animations; repeating, granted, but at least not distracting. (The Euphoria engine is a step in the right direction. Youtube search for "euphoria physics" for examples.)


Fluid dynamics: still waiting for processing power to become available. Bioshock and Crysis had relatively pleasing water & fire (I liked it), although to be objective both still look precomputed. (Note: I've seen very little of Bioshock DX10 fire, if it's any different, so maybe I'm wrong.) Smoke is improving, but you generally have one of two types. The Counter-Strike type, which is a sort of haze but seems restricted to a fairly strict area and doesn't react to motion/wind; and the layered texture type like the rocket at the end of HL2:EP2, which looks pretty from a distance, but try looking up and rotating your view (the layered textures rotate with you). STALKER: Clear Sky is trying to add interactive smoke, so I'll wait and see what it's final product looks like.

Lighting: probably the area that will have some of the most noticeable improvements in the near future. Dynamic shadows are improving quickly. Also, "screen space ambient occlusion" looks like an area for great rewards. Crysis already uses it, so hopefully we'll be seeing more and better versions of it. (A straightforward example: http://www.projectoffset.com/blog.php?id=83) That Project Offset screenshot has some small(ish) flaws (like around the hand), but I'm still impressed!


Other random _types_ of things I'm looking forward to:
* Deformable models: if I shoot a barrel with a shotgun it'd be nice if I could at least dent it, if not see the light on the other side. (This isn't the same as shooting a wooden barrel and having it break into new pieces.)
* Procedural icicles that can grow, melt, drip, and be broken off.
* Better interactions with snow (good footprints, puffs when hit, builds up on objects INCLUDING CHARACTERS, etc).
* They're already really improving building destruction. (Related to deformable models.) Terrain forming is also still a work in progress.


Ugh, too much typing... Probably some mistakes in there.

 

[pmv]

Originally posted by: Drift3r
Actually the AI is there for stronger opponents but you folks need to understand that game developers make games for everyone not just a small segment of hardcore players looking for a challenge. A friend of mine who works for a gaming company told me once that they could easily add in AI that learns and adapts to players gaming tendencies. Of course the problem would be that over time this type of AI would quickly frustrate and put off many potential customers who don't have the time or patience to always keep things varied in a game in order to beat this AI. Most PC game companies always look for ways to reach and target the most people when they make a game thus you'll almost always have AI that can be beat by anyone with a IQ at or above 70.

That certainly isn't the case for strategy games. Few strategy game AI's employ any actual strategy, they just try to exploit their higher boredom threshold and out-micromanage you. And they cheat of course. Playing something like Civ against the AI or against human players are totally different experiences.

At most surely your arguments apply to bots in simple FPS deathmatch type games?

Oh, and of course if adding good AI were so easy as you suggest, developers could easily include different levels of AI, changing difficulty level by changing AI level instead of just changing damage ratios or cheating levels. Thus catering for all levels.

 

[Drift3r]

Originally posted by: pmv

Originally posted by: Drift3r
Actually the AI is there for stronger opponents but you folks need to understand that game developers make games for everyone not just a small segment of hardcore players looking for a challenge. A friend of mine who works for a gaming company told me once that they could easily add in AI that learns and adapts to players gaming tendencies. Of course the problem would be that over time this type of AI would quickly frustrate and put off many potential customers who don't have the time or patience to always keep things varied in a game in order to beat this AI. Most PC game companies always look for ways to reach and target the most people when they make a game thus you'll almost always have AI that can be beat by anyone with a IQ at or above 70.

That certainly isn't the case for strategy games. Few strategy game AI's employ any actual strategy, they just try to exploit their higher boredom threshold and out-micromanage you. And they cheat of course. Playing something like Civ against the AI or against human players are totally different experiences.

At most surely your arguments apply to bots in simple FPS deathmatch type games?

Oh, and of course if adding good AI were so easy as you suggest, developers could easily include different levels of AI, changing difficulty level by changing AI level instead of just changing damage ratios or cheating levels. Thus catering for all levels.

By easy I meant they could fund and pay for really expensive AI and programmers that would get the job done. Though the more complex a game is to code for the higher the production costs and more people are needed to be involved. Adding more complex AI is easy but then you have to pay people $$$$$$$$ to code it for you and it would increase the cost of the game tremendously. So what do developers do? They use simple tricks as you stated to give their simple but cost effective AI a fighting chance at winning while keeping development cost reasonable enough to turn a hefty profit on their game. Decisions by most game companies when developing and creating new games are made to reach the greatest audience and turn a profit and sometimes that means focusing in certain areas (like graphics) and leaving other areas (like strong solid story, really solid AI code, etc...) at a bare minimum. Not to mention the fact that they have to also balance computing power for the things they want to do in game without turning the game into a slide show for most consumers.

 

[Frodolives]

Graphics can always evolve, but they will forever be an abstract. It's the interface that is the next frontier.

 

[pmv]

Originally posted by: Drift3r

By easy I meant they could fund and pay for really expensive AI and programmers that would get the job done. Though the more complex a game is to code for the higher the production costs and more people are needed to be involved. Adding more complex AI is easy but then you have to pay people $$$$$$$$ to code it for you and it would increase the cost of the game tremendously. So what do developers do? They use simple tricks as you stated to give their simple but cost effective AI a fighting chance at winning while keeping development cost reasonable enough to turn a hefty profit on their game. Decisions by most game companies when developing and creating new games are made to reach the greatest audience and turn a profit and sometimes that means focusing in certain areas (like graphics) and leaving other areas (like strong solid story, really solid AI code, etc...) at a bare minimum. Not to mention the fact that they have to also balance computing power for the things they want to do in game without turning the game into a slide show for most consumers.

Well that was my point, that to include decent AI now would require huge amounts of work (and I'm not convinced they could even do it at all in all cases) because the off-the-shelf-tools and _general_ techniques aren't there, nor is there any suitable hardware acceleration. In contrast to graphics in which progress is clearly being made continuously as recent work builds on earlier work to allow more to be done with the same effort, both in hardware and software.
Graphics is the easier problem to provide general solutions and tools for, so that's what gets the attention.

Better stories is the other part though - surely one of the things that made HL1 a classic was that they bothered to employ an actual writer to develop the storyline (did quake 2 do that? somehow I doubt it, or if they did it was a very bad writer). Even then, the best PC game stories just rise to the level of OK genre fiction, its still hardly literature.

I think its true though that better graphics do enhance stories, by providing more atmosphere. And I reckon the modelling and texture work in games is under-rated as art, those guys often do a fantastic job at providing a consistent style and theme to a game (e.g. the futurist look of Sin or the stylised horror imagery of Clive Barker's Undying). Graphics isn't just about creating a perfect, naturalistic virtual reality imitation of the real world, its got to have some artistic stylisation in it as well.

 

[sandorski]

Originally posted by: trelin
Disclaimer: I'm not complaining here. Also, I'm not an animator, texture artist, modeler, or whatever. I'm just a random joe that notices too many details. There's also a lot I've never seen, I don't claim comprehensive knowledge, this is just off the top of my head.


Textures: Seems easy enough but my biggest pet peeve with new games are poor quality textures. Take a game like Max Payne 2. Very low poly stuff, but the textures were so clean/crisp it never bothered me. I remember one scene in particular I got killed just admiring the quality of a large billboard inside a hotel/office. Keep the textures clear and work up all the other effects from there (I'd rather have lower detail textures than high detail/high compression artifacts). A lot of this goes down to good art direction dictating how well all the textures work together. For example, Dreamfall didn't have particularly high detail textures, but they worked well together.

Biological/organic animations: my biggest wish is for randomized animations, or at least more versions of each animation. It gets tiring seeing exactly the same skeletal changes every time a character jumps/gets shot/whatever. They're just so... stiff, like gears with strict articulation, and no sense of weight. A good comparison is to synthesized music: if the beat is absolutely dead-on every time, it gets fatiguing; reduce the "accuracy" of the synthesized beat and it begins to sound more natural. A good example of stiff animations is Halo (1,2,3); when a person jumps/falls it always looks exactly the same, you can literally see where the animation sequence starts and stops. Also, when a standing person looks round, their feet don't move, like they're floating. This is an area which TF2 has done quite well, all the characters have pleasantly fluid animations; repeating, granted, but at least not distracting. (The Euphoria engine is a step in the right direction. Youtube search for "euphoria physics" for examples.)


Fluid dynamics: still waiting for processing power to become available. Bioshock and Crysis had relatively pleasing water & fire (I liked it), although to be objective both still look precomputed. (Note: I've seen very little of Bioshock DX10 fire, if it's any different, so maybe I'm wrong.) Smoke is improving, but you generally have one of two types. The Counter-Strike type, which is a sort of haze but seems restricted to a fairly strict area and doesn't react to motion/wind; and the layered texture type like the rocket at the end of HL2:EP2, which looks pretty from a distance, but try looking up and rotating your view (the layered textures rotate with you). STALKER: Clear Sky is trying to add interactive smoke, so I'll wait and see what it's final product looks like.

Lighting: probably the area that will have some of the most noticeable improvements in the near future. Dynamic shadows are improving quickly. Also, "screen space ambient occlusion" looks like an area for great rewards. Crysis already uses it, so hopefully we'll be seeing more and better versions of it. (A straightforward example: http://www.projectoffset.com/blog.php?id=83) That Project Offset screenshot has some small(ish) flaws (like around the hand), but I'm still impressed!


Other random _types_ of things I'm looking forward to:
* Deformable models: if I shoot a barrel with a shotgun it'd be nice if I could at least dent it, if not see the light on the other side. (This isn't the same as shooting a wooden barrel and having it break into new pieces.)
* Procedural icicles that can grow, melt, drip, and be broken off.
* Better interactions with snow (good footprints, puffs when hit, builds up on objects INCLUDING CHARACTERS, etc).
* They're already really improving building destruction. (Related to deformable models.) Terrain forming is also still a work in progress.


Ugh, too much typing... Probably some mistakes in there.

All these things will eventually get solved, but to a large extent these issues are Hardware Limitations. Some Developers are lazy or too rushed to address these issues, but in their defense making multiple Animations, Deformable Objects , and many other things takes a lot of time to implement correctly.

 

[darkswordsman17]

Originally posted by: sandorski

All these things will eventually get solved, but to a large extent these issues are Hardware Limitations. Some Developers are lazy or too rushed to address these issues, but in their defense making multiple Animations, Deformable Objects , and many other things takes a lot of time to implement correctly.

Yeah, and that stuff tends to be a negative if its not done correctly. Look at ragdoll physics, where it was heralded as being so realistic, but it ends up being less realistic in the end (and leading to some very silly effects).

 

[trelin]

Originally posted by: sandorski
All these things will eventually get solved, but to a large extent these issues are Hardware Limitations.

Whoops! I thought your response was a little odd until I realized that my post missed the OP's question completely. (I'd just been having this exact discussion with a friend and braindumped some of my thoughts.) I didn't mean to put those issues up as limitations on where graphics are heading, but merely examples of some upcoming improvements that I thought were neat. Didn't mean to change the course of discussion, sorry.

 

[cyclohexane]

Originally posted by: GundamSonicZeroX

Originally posted by: ConstipatedVigilante
We still need decent virtual reality, man. Now THAT would be awesome. WTF would happen when it crashed, though?

You would be lying on the floor motionless seeing an error.
"Would you like to send an error report?"

qft