Rasterman - you shouldn't need prior experience with an enemy to be able to kill it. Did you die every time you encountered an enemy type for the first time? Of course not. There's some basic things you can learn from examination. For example -

This guy has a spear. He probably has a pretty decent melee range; better be careful.

Hrm, the floating girl has created a magic-seal-looking thing and is waving her arms around. I bet she's going to cast a spell; I should be ready to dodge.

That guy with the huge sword just did a huge windup. Better get out of here!

That panther just crouched back a bit; I bet he's going to leap forwards soon.


Now, enemies shouldn't completely broadcast everything about them - you don't need to dodge every attack, after all, to win a battle. But a good enemy design will provide hints as to what it's capable of, as well as sufficient clues that an on-the-ball player will be at least partially prepared for what's coming. Personally, I find the first encounter with a given enemy type to be the most interesting, because I'm learning how the enemy fights at the same time that I'm trying to defeat it.

As a random example, take the Mega Man series' bosses. The first time I fought, say, Crash Man (from Mega Man 2) was a much more interesting and close fight than the last time. Crash Man, like most Mega Man bosses, has a very basic pattern, but it takes a while for the player to find the corresponding counter to that pattern. Until you've figured out the pattern, he can be very hard to beat! You don't know where the safe points are, so you have to react quickly when he decides to jump and launch a bomb at you. Whereas once you've learned the pattern, you know exactly when he's going to launch that bomb, and by the time he does you're already well out of the way. The fight is now much less interesting, being just a matter of exploiting Crash Man's AI in a completely safe fashion.

Now imagine if you had to fight Crash Man not just twice (being a Capcom game, of course you have to fight every boss twice!), but several times in each level. What was originally a fun fight, and then became largely a matter of practiced timing, is now tedium. That kind of problem is why varied enemies are worthwhile.

Quote:

Original post by Kazgoroth I've noticed a distinct lack of variety amongst the evil we encounter during our many adventures. ... ... My question to you, good forum members is if you see any merit in attempting to provide more variety to our game worlds by solving this problem. ... ... There's also the argument to be made that producing variations on our enemies creates a significant amount of additional work for the art department Your thoughts?

*snipped* to the bits i'll reply to :)

Yes most games just create their basic enemy types and re-use them everywhere they can. That's usually because of perceived (and sometimes real) technical limitations though occasionally it's a deliberate stylistic choice to stop you thinking about those enemies as individual things.

This problem should and in some demo's I've been privy too it has been solved. One of my favourite that you can view online is by Ken Perlin (google for his website) and it's his animated face using noise. You could have 10,000 of those faces throughout a world and with varying expressions you could give them all a good dose of individuality before you even got onto colouration, shape of the face or anything else.

The art problem is one of design and pipeline. If your artists are given the tools and means to create figures that can be easily modified using a number of parameters then they should only need to create the same number of models as they originally planned. After all we're all human (ok so maybe that's a HUGE assumption on here :D) but we don't vary too much. Certainly most of our variations could be described and parameterised.

Developing the tech to do all of the above, and creating tooling for the art pipeline is a large job and thats mostly the reason that you dont see it done in games. There's a certain amount thats understood and theres a very large risk in deviating from that as the deadlines get tighter and tighter with each game. Unless it was one of you Unique Selling Points/features or that game *required* it in some way (such as the one I'm on) you're unlikely to see it done.

Finally there's the cost of this kinda of tech. If you have one model with one animation then you just blat it to the screen as many times as you need it. If you have one model with unique modifiers then you need some clever way of modifying each and every one that you need. Perhaps thats on-the-fly during rendering or cached somewhere in memory. Either way those approaches costs you resources, rendering performance or memory.

So usually when the idea comes up someone will say; "but we need that memory for X feature which is more important" or the people responsible for the gfx will say; "yeah we can do that but we'll have to drop M, N and/or O features because of blahblahblah". Those reasons may or may not be entirely valid but it's hard ot tell at the start of a project when you've got to estimate your initial workload for the next 18 months :)

In short: Yeah I think it can be done, and you'll see much mroe of it for this generation of consoles / PC games due to people investing in the tech to do it all. Though it might not knock your socks off in amazement just yet.

Andy

In an era where many games have customizable characters, you'd think someone would come up with procedurally created monsters.

<RANT>
One of my biggest peeves is when monsters are designed with wounds and then "cloned" so that every monster of that type has the same wound. I physically cringe when I see this in a game. A good example of this would be the fat zombie monster in Doom 3... the worst I've seen was when I played "Ravenloft" while in college (dreadful game, btw) and my character was being attacked by an army of zombies, each one missing the same arm. Did they REALLY have to have one-armed zombies?!?! Wouldn't it be better to give 'em two arms so the game would look LESS retarded? GAH!
</RANT>

*deep breath*

Quote:

My question to you, good forum members is if you see any merit in attempting to provide more variety to our game worlds by solving this problem. Obviously back when Super Mario Brothers was new there were some pretty extreme limitations on memory, but this is a problem that is significantly reduced for us in the current day-and-age.

Um, I don't exactly see this as a problem...Its like saying Chess would play better if every piece like the pawns both looked and could move/attack differently. That kind of stuff can ruin gameplay.

It's to be handled on a per-case basis. Obviously if you want your creatures to look the same, make them look the same. The point Kazgoroth was making is that it doesn't always make sense. In fact, it rarely makes sense. In the case of goombas, I don't see a problem with them all looking the same. They're little mushroom people, and I wouldn't expect them to have much variance (though it might be nice in a modernized Super Mario Bros.).

The Sims and Oblivion are taking this in the right direction, but you have to admit it hasn't gone far enough. While the people in Oblivion have different faces, their builds are nearly identical even between races (as all races use the same base model and texture for both body and face), and they lack blemishes or other physical idiosyncrasies (scars, tattoos, moles, etc.) that really add to a character's identity. Also, none of the monsters are different; every wolf looks exactly the same as every other wolf.

The biggest problem I see with the clone philosophy is a lack of emotional investment. It is impossible to identify with a creature that looks like every other creature of its type. This is difficult even when they try to give a particular creature some personality, such as Skuz in Stonekeep (he still looked like your run-of-the-mill Sharga, even if he talked different). The key to making a character lovable is to make it unique in a manner your audience can appreciate. And that's another thing to keep in mind: players are not the audience, they are your audience.

So, we're moving in the right direction, but clearly we have a long way to go. I think procedural differentiation is definitely something we should pursue as an industry, and I think it will really take off once we're finally able to render a world that looks almost entirely real (GF-27000 ought to do the trick), because part of making it look real will be to avoid the clone problem.

Personally, I too am annoyed by the lack of variation, especially with games that offer a limited selection of appearances, and they all suck (as was the case in previous Elder Scrolls games and is still the case in most other games, most notably the Guild 2 — seriously, why can I not wear a hat?).

There's a good reason why we tend towards iconic images: it's how our brains are designed to work. Difference is what attracts us. It's what draws our eye. When we learn to pigeonhole "Friend" rapidly, we can use more of our energy trying to spot "Foe" and our brains are optimised for such processes.

In an RTS -- much as in war generally -- the need to be able to spot, at a glance, whether the guy you're looking at is on your side is far more important than what his life story is. Hence the preference for uniforms with garish colour schemes during much of our species' warmongering. (It's only relatively recently that we've cottoned on to tricks like camouflage.)

The argument in favour of these wild colours is simple: It's one thing to see me and my army coming, but quite another to be able to do something about it. In the main, battles tended to be one-sided affairs. (We hear far more about the close-run fights which could have gone either way, because they make for much more interesting stories, but they're not all that common in history. A successful strategist would always prefer to have overwhelming force on his side and make damned sure that's what he's got.)

The clone-like nature of enemies in FPS games is a holdover from the days of old-school shooters, when aliens would fly towards your ridiculously overarmed spaceship (which, inexplicably, appears to lack brakes or even a reverse gear), in simple, easily memorised formations. All FPS titles owe a tip of the hat to Wolfenstein 3D, which was basically a 3D take on Atari's "Gauntlet". These early games recycled the same old sprites because of simple memory and VRAM constraints. FPS titles have been incremental evolutions from these ancestors.


The requirement to economise on graphics data is dependent on a simple ratio: that of graphics storage to system power. The more data you need to shunt around, the more power you need to be able to do that. Offloading work to graphics cards helps, but those cards have defined VRAM limits and game developers need to ensure their games will run acceptably on a wide range of platforms if they want to see a profit. Because of this, the granularity of the raw graphics data has been quite poor, with characters essentially being constructed from a very small kit of assets.

As RAM increases and processing power follows suit, the need to produce such large lumps of pre-digested graphics data for the computer in a ready-to-render form reduces. The kit of assets available will increase and in-game entities -- from characters right through to the environment itself -- will be built from increasingly granular databases.

Within ten years, the notion of hand-building a few, copy-and-paste cannon-fodder enemies will seem positively prehistoric. Assets will become increasingly procedural. (Except in cases where simulation of a real-world location is necessary, for which downloading ready-made "content clip-art" may well become the norm.)

This will happen. It's inevitable. The existing asset pipelines simply won't scale without a move towards supporting increasingly granular asset design.

(That said, I don't think 2D puzzle games are going to disappear any time soon, so there'll probably always be a place for traditional pixel-pushers.)

stimarco... Didn't know you knew about this place...

I personally think that enemies should be intelligent in different ways. This could be done by 'brute force' programming it in, teaching the computer various strategies to choose from, or just establish a goal for the computer and let it do it however it wants. This last way is of course somewhat more difficult, but I have played games that do incorporate this (or something VERY much like it), and they work very well. Game play is more tactical and intelligent; each player, the computer and the person choosing how to attack or defend. I don't just mean first person shooters, either.

If you just want physical variety in a game, all the developers have to do is add a random function. If, for example, you wanted players wearing different color shirts, a random color could be painted on them.

Quote:

Original post by MSW Quote: My question to you, good forum members is if you see any merit in attempting to provide more variety to our game worlds by solving this problem. Obviously back when Super Mario Brothers was new there were some pretty extreme limitations on memory, but this is a problem that is significantly reduced for us in the current day-and-age. Um, I don't exactly see this as a problem...Its like saying Chess would play better if every piece like the pawns both looked and could move/attack differently. That kind of stuff can ruin gameplay.

Pawns are meant to be identical. Human characters in games are notionally meant to be humans, and humans are not identical.

I think clones will increasingly be a problem for suspension of disbelief as graphical realism increases: particularly with faces.

As humans, we associate people's faces with people's identities, and the more detail we have about a face, the more certain we are of the identity of its owner.

In older games, there was insufficient graphical detail for our brain to make the connection between this cookie-cutter enemy and the last one with the same face. We might notice a family resemblance if we thought about it, but our hindbrain is not telling us "you have already killed this person".

Generally, only unique plot-specific NPCs had sufficient detail that you'd notice if they had a clone. Since there were only a handful of them, it would be relatively easy to ensure they weren't identical.

One an example where this didn't happen is in Neverwinter Nights: some of the unique NPCs's face portraits were available for use as character portraits. The effect was quite jarring when you met an NPC who your hindbrain mistook for a PC you'd played before.

In modern games, though, there is a huge amount of graphical detail even for "grunt" characters. Enough that, if we get a good view of a grunt's face, it may seem odd if another grunt with exactly the same face appears a little while later. The last game I played where this was particularly noticeable was Jade Empire, where every male Lotus Assassin appeared to have exactly the same, highly detailed, face.

Worth noting is the fact that this only applies to human faces. Half-Life 2 gives each human a different face, but it's not at all odd to us that all the combine wear the same helmets, or that all the Vortigaunts are identical.

In rare cases where an alien is visibly different, it is usually a big difference. If you've played Halo 2, you probably remember the half-jawed covenant elite commander, Rtas 'Vadumee. For a player to identify Rtas as a unique character, it was necessary to remove half his face, but consider how grotesque that would have looked if Rtas was a human with half his jaw missing.

The actual technical challenge of generating different faces is pretty much solved. You can generate faces on the fly using a set of morph targets for basic face shapes. Many of those faces look pretty stupid, so the game's developers would probably still want to select specific faces from those possible, but the face generator itself can provide rigging suitable for lip-synching and facial expressions. You can have thousands of distinct faces for a cost only marginally greater than modeling 20-or-so facial archetypes.

Quote:

Original post by Geometrian stimarco... Didn't know you knew about this place...

I'm a generalist with over 20 years in this industry. I've been a designer, programmer, artist, animator, consultant and even a technical writer. I've now returned to the industry (after a few years' sabbatical in education) as a one-man indie designer / developer.

Most successful game designers will be generalists, I think. It's an inevitable side-effect of the role.

Quote:

I personally think that enemies should be intelligent in different ways. This could be done by 'brute force' programming it in, teaching the computer various strategies to choose from, or just establish a goal for the computer and let it do it however it wants. This last way is of course somewhat more difficult, but I have played games that do incorporate this (or something VERY much like it), and they work very well. Game play is more tactical and intelligent; each player, the computer and the person choosing how to attack or defend. I don't just mean first person shooters, either.

That's an AI issue rather than a visual one. Nothing wrong with that, but I don't think Kazgoroth was complaining about this. (Besides, some players like a certain level of mindless blasting.)

Quote:

If you just want physical variety in a game, all the developers have to do is add a random function. If, for example, you wanted players wearing different color shirts, a random color could be painted on them.

And this is the real problem. Changing the colour of clothing isn't really enough today. For games like Tom Clancy's Rainbow Six series, where characters are covered entirely in military gear so you can't see their faces anyway, your solution wouldn't work; armies prefer to purchase their kit in bulk, so the similarity in looks between characters is intentional.

However, in -- for example -- an RPG, it is not a requirement for everyone to look exactly the same barring some minor cosmetic clothing changes. Unless we're talking about an unbelievably in-bred population, the visuals simply won't be immersive. The player needs to be able to suspend their disbelief. As long as graphics were simplistic and at the iconic / cartoon end of the scale, recycling the same graphics was acceptable. However, developers -- okay marketing people -- are hyping realism and that means you have to ensure characters look realistically different too. Kazgoroth's point is a valid one.

I remember when racing games used to recycle the same sprites for all the cars. (Anyone remember the original Atari "Pole Position" game?) Nobody would ever accept such a game today. Why should the FPS genre be an exception?

There's this Turkish indie game called Mount & Blade (still in beta, but well worth the $18), that does this in some ways. There are various classes of units in that game, but you never know until you get out on the battlefield exactly what you'll be facing; sometimes a guy will throw down his crossbow and pull out a mace, other times he'll be carrying a sword. Also, there's a pretty elaborate face creator, so just about every npc has a unique head and face.

I hear from the forums on their website that people run it on pretty archaic hardware without much trouble, and with full effects, my 2 ghz, 2Gb ram laptop only bogs down when there's over 100 npcs in the line of sight.