eigenface

There is a "good" way to do flying AI within the limits of doom 3, floor-based pathfinding, because the cacodemon did it. The situation is more complicated now, because the dark mod AI is more complicated (vision, patrols, alerts, etc), but that shouldn't make a difference as far as using the doom 3, floor-based pathfinding is concerned. Our flying AI *should* be able to do something very similar to what the cacodemon did, that is, create a convincing illusion of true 3D flight, using the doom3, floor-based pathfinding plus dynamic changes in hover height. Does this sound right?

Thanks for the info, demagogue. To clarify, is the following correct: There were 2 kinds of AI code in doom 3, engine AI source code (C++) and AI scripts (a C++ based scripting language), and by contrast, there is only 1 kind of AI code in dark mod, engine AI source code (some of it was already present in doom 3, and more of it has been added by the dark mod team)? I want to make a reactive enemy that has vision, alert states, and patrol routes (albeit hovering above them) the same as the existing dark mod AI, so it sounds like jury-rigging a will-o-wisp would not be enough. Does that sound right? What would you recommend: should I use the scripting language, or would I be better off modifying the dark mod source code? I have no problem modifying the source code, but then the flying enemy would not be available until the next dark mod release. The scripting language sounds like the better option to me, but then again I have no idea what would be entailed in getting it to work. Bikerdude, are you talking about a flying bird programmed with the scripting language? Let me put the question this way: does it sound like a reasonable task to try to get a basic flying enemy working (vision, alerts, patrols, etc) by reworking the cacodemon script to function with dark mod? Or does that sound like a fool's errand?

Thank you for a very informative answer. Actually, the flying AI I have in mind would work well as a variable-height, will-o-wisp variant. I'm a programmer who's worked in Java and AS3 (flash games), so this may be within my grasp. Am I to understand you've removed the AI scripting in favor of pure C++ AI baked into the engine? This would mean the only way to make (significantly different) custom enemies is to edit the mod source code.

How feasible is it to make a flying AI? Can they follow patrol waypoints like walking AI? Doom 3 had them, but their AI was much simpler to being with. Is it possible, for example, to put the cacodemon in the dark mod? Not that I want to, but I may want to use the cacodemon as a starting point for putting together a flying enemy. Or is the dark mod already so different than doom 3 that the cacodemon is no longer functional or relevant as an AI?

This is without a doubt one of the best missions I've played - old-school horror. Gave me goosebumps, like Inverted Manse all over again. I found one bug, climbing on top of the upright sarcophagus in the crypt, at the end of the hallway that leads into the chamber with the book. If I keep close to the wall, and then jump up and mantle onto the sarcophagus, I swing right through the wall during the mantle animation, and I can see outside the map for a moment. This is easy to reproduce, and I tried to take a screenshot, but I couldn't get it to come out right. This might be possible in other missions - it doesn't seem like it would be unique to a certain map, but I'm not sure. I looked at the mantling code, and as far as I can tell, it checks whether the path is clear for you to climb up into, but it doesn't check whether the swinging-side-to-side part of the mantling animation has room to stay inside the walls. Does this make sense?

I got the same map issue with the scrolls on top of the bookshelf in the scriptorum - the inventory freezes as soon as I pick one up. This is after updating to 1.08. In addition, I think there's something wrong with the update process, because every time I go to the new mission menu, multiple missions give me the "clean-up" option. I click it, and confirm in the dialogue box, for every mission, and the clean-up option still appears for all those missions, even after I restart the mod. I also get weird visuals with water splashes - everything behind the splash gets washed-out and white-tinted around the edges. This also sometimes happens with fog - I can see square edges and boundaries between different volumes of fog. On the plus side, I no longer need RadeonPro to prevent the world from rendering upsidedown. win7 x64 ATI Radeon 5970 with updated drivers

I just want to say 1.08 rocks! Congratulations and thank you to all the folks who worked so hard on this release, and all the releases that lead up to it, not to mention the amazing FM authors. I feel like the mod has reached some kind of turning point. I started a few missions in the past, but something always kept me from sticking with them, something always broke me out of the game world. But now I've gone back and replayed the same missions, and they were beautiful, captivating experiences from start to finish! The dark mod 1.08 feels like a polished, AAA-game, just as good and better than the original thief games that inspired it. Now that the mod is fully capable of doing justice to their visions, the number and quality of fan mission authors will continue to go up and up. It's finally all coming together. Dark mod is absolutely the best mod out there, and everyone who's worked on it must be extremely proud. You guys rock!

An even better example: http://gamejolt.com/freeware/games/adventure/which/1523/

"sketch post-process shader" Someone did it! Seems to work pretty well. Give it a play. http://www.superfriendshipclub.com/forum/viewtopic.php?f=12&t=117

http://en.wikipedia.org/wiki/Feature

Regardless of whether you're for or against the push, what do you think about adding falling and landing animations, for all those miscellaneous situations where an AI could end up falling (a trapdoor opens under them, the bridge they're on collapses, etc)? Would this mess-up their path-finding, or would they realize where they are once they land and continue?

A push move would certainly add new gameplay possibilities. I'm thinking that's not a bad thing. I can't imagine it would break many plans - it does no good unless the AI is on the edge of a precipice (or next to a dangerous object, fire, moving gears, etc.) The mapper pretty much has to set up a situation where the push would be useful. Otherwise, it just alerts the AI. I'm thinking once they're alerted, pushes would no longer affect them (they're ready for it and bracing themselves for an attack.) But even if it did still affect them, it's not a good tactic since they'll be slashing you to ribbons. I don't see why you think the AI should go ragdoll when it falls - I don't go ragdoll when I fall. I guess I sort of wave my arms for balance and bend my knees so I can absorb the impact when I hit the ground - a falling animation. "Next thing people want is the ability to hit an armed AI in the face and steal his sword.. Or maybe take human shields and fire arrows behind them?" I guess you're going for some kind of slippery slope fallacy? Picking certain weapons from unaware guards could be neat, though - I thought it worked well with picking the pagan shamans' wands in Thief 3.

Actually, is there a crouching animation already? I don't even know. A crouching animation could be used for a variety of purposes, including landing after a fall.

This sounds like a neat feature. I think a generic falling animation and a generic landing-crouch animation are all you'd really need, and they could be used in several other places - what happens presently in dark mod if a trapdoor opens under and AI or the bridge he's standing on collapses? Something not very natural-looking, I'd wager. Eventually, it would be good to have falling and landing animations, even without the push feature. It's only slapstick if you design it that way; on the contrary, a person falling to their death can be a pretty disturbing sight. I'm guessing the logic would be something like run a check of the distance from the AI's feet down to the floor (either periodically or only on certain events, I dunno how the dark mod works), and if it's greater than some threshold, they go into the falling animation. If they die when they hit the ground, they become a ragdoll, otherwise they go into the landing animation. Then if you wanted to implement the push, all it would do is move the AI a small distance (you could probably get away with using the normal walking animation if the distance is short enough - no need for a getting-pushed animation.) If they didn't fall off anything, or if they only fell a short distance (less than the threshold) like being pushed off a single step of a staircase, they wouldn't go into the falling animation at all, they'd just get alerted and turn around to attack you. Now, getting pushed down a staircase would be a whole other can of worms...

Fidcal, I agree the readme should let the player know the mission is breaking a common convention, otherwise they'll certainly expect the same kinds things they've played 100 times before. In fact, it should probably state this twice, once explicitly ("Not all locked doors are openable") and again in the story ("This is a wealthy part of town, and the denizens know the first thing a rich man should buy is a good lock. I'll probably never get inside most of these houses.") "In any event, even if having all doors frobable is not bad design but simply 'different' it still might not be so enjoyable for the majority." That seems premature, because the idea is completely untested. Not sure if the majority are scourers or what. It might be worth thinking about the question: Does the majority even want the illusion of a large world without boundaries? I assumed the answer was yes, but I could be wrong. Maybe most people want to find the limits of the artificial environment, so they can make sure they didn't miss anything. To me, running into contrived boundaries always kills the atmosphere, fake doors remind me the world is fake, etc.

"If a door is interactive, it's reasonable to assume it can be picked or unlocked with a key." It's reasonable to assume this in the artifical world of the video game, because virtually every map made so far follows this convention. It's completely unreasonable to assume this in real life - there are interactive doors everywhere which you'll never be able to unlock (in fact, all real doors are "interactive", except fake doors on a movie set or something.) The question is, is there any gameplay or atmosphere advantage to be gained by breaking with this convention? A convention is not the one right way, nor even necessarily a good idea (I bet there are several conventions of dumbed-down console FPS games which you don't like.) As I've argued, I think interactive, unopenable doors have the advantage of making the video game world seem less artificial, by creating the illusion of a large world without physical boundaries - from the player's POV, there's always the chance they'll find another key and unlock a new area. I think I understand now why you've mistaken this for bad design. It's probably my fault for not making this clear in first place: there are NO non-interactive doors. You're imagining a map with some scenery doors and some locked, unopenable doors. Yes I agree, that would be misleading and frustrating to play, but that's not what I was suggesting. Every door in the map which would normally be scenery is now locked (with no key or ability to pick.) So you're not misleading the player - the whole point is you're not leading the player at all, at least not with "special" (interactive) doors. When all the doors are interactive, navigating the map is no longer about figuring out which doors are real and which ones are scenery in this artificial environment. Now there's no point to thinking about the artificial boundaries, because you'll never know them when you find them.

"If the player encounters a door that is indistinguishable from other doors and yet it cannot be opened that's bad design." Yes yes, I already know many people like to assert that, but I have yet to hear any counterarguments to my above reasons explaining why it's actually good design. This seems like one of those things "everyone knows", that is, one of those things everyone takes for granted and doesn't question. Also, "indistinguishable" is misleading - the door on the Cathedral is in fact distinguishable from the door on the random house, and the player should know to try the key they took from the Builder on the Cathedral instead of on the random house.

I think the Planner versus Explorer dichotomy makes a lot of sense, Sotha. And it brings to light a problem I have with most thief and dark mod missions, nay, most video games period. My problem may be due to my own obsessive nature, and I'm curious as to whether anyone else thinks the way I do. At heart, I'm a planner - I want to carefully come up with a winning strategy, then get in and get out, like a real thief. Exploring every last nook and cranny in a mission is an immersion-breaking process in and of itself, because it has no analogue in real life. In real life, it's ridiculous to think about "exploring everywhere" and "stealing everything", because there's always, always more, way more than you could ever possibly explore or steal. And as you steal more and more, the odds of getting caught increase and increase, eventually approaching 100%. In real life, you want to achieve your one narrow objective quickly and with as little risk as possible. Finding the edges of the map and hunting down every last piece of loot are processes inherent to the artificial nature of your environment, and they draw attention to it. In practice, I'm an explorer. Why is that, when I've just spent the last paragraph trashing that gameplay style? Because in the back of my mind, I know the game world is artificial. I know it has physical boundaries and a limited number of loot items and enemy AIs. And I know it was all made for my entertainment. Perhaps I don't have the discipline to plan an execute a "get in, get out" strategy when I know I can just use brute-force and inevitably win by attrition (I'm counting blackjack KOs as "brute force".) But I think there's more to it than that. Let me explain. In the back of my mind, I know everything in the game world was put there because the author thought it would contribute to the experience in some way. And I want the full experience. I want to appreciate all the author's work, and wring every bit of fun out of the mission. Ironically, that ends up being less fun, because I won't stop searching until I run into the game world's boundaries and limits, and then immersion is broken. In fact, immersion is broken even before I find the boundaries, because I know they exist and I'm actively looking for them the whole time I'm playing. I'm thinking "It that a frobbable door? Nope, it's a scenery door. Is there a way up to that window sill? Could be a hidden loot item up there. Can I get to the area behind that grate? I don't think so, I bet it's one of those unreachable areas they put in to make the mission feel bigger." The most immersive part of the mission is usually near the beginning, when I haven't started worrying about the boundaries, and for a little while I have the illusion of endless possibilities. Does what I'm describing sound familiar to anyone? Sandbox or "open world" games attempt to address this problem by giving you a larger environment than you can possibly explore (or at least, larger than you probably will chose to explore.) The problem is, game development resources are still limited, so the open world ends up being a whole lot of boring interspersed with little nuggets of interest. The environment may be vast, but that usually just means the interesting events are spread more thinly. Obviously, this reflects my own negative opinion of sandbox games. I know the idea is that interesting events are meant to be procedurally generated, emerging from the interactions of the AIs in different situations, without the need for each event to be individually designed. I'm of the opinion that designed events are usually more interesting, at least with the current state of AI. Closed-corridor games-on-rails like Half Life 2 are often more immersive than open-world games, because the rails games do a better job creating the illusion of a large world than the open-world games do fleshing out and populating an entire large world. I know there are people who find the process of wandering around an open-world environment immersive in and of itself. Personally, I end up doing something similar to what I do for thief and dark mod, only instead of searching for the boundaries of the mission, I'm going down an estimated mental checklist "Have I done everything interesting I can do in this sandbox?" or "How much more wandering around am I going to have to do before I stumble across a new interesting event, and is it worth it?" I think thief-style missions are the perfect compromise. They're not closed-corridors; there's plenty of room and alternate routes for you to wander around and encounter emergent AI behavior. And they're not open-world; they're small enough the "density of interest" can still be pretty high, even with the modest development resources of fans. In my opinion, the problem is this: not enough priority is placed on creating the illusion of a large world without boundaries or limits. Part of the problem is players have become savvy - they know there will be unreachable areas put in to make the world feel bigger. Nonetheless, there are simple things authors can do to take away or greatly reduce the player's awareness of the boundaries of the environment. For example, add unpickable locked doors with no key. I know this is anathema to some people. I've heard people say "Every locked door either should be pickable or it should be possible to find the key." Clearly, they were Explorers. Of course every lock has a key, but maybe it's behind the locked door. Maybe it's miles away on the other side of town with it's owner. Maybe it's been washed away from his mugged corpse in a gutter somewhere and carried down a storm drain. Just because the key exists doesn't mean you have a real chance of finding it. The real world is full of locked doors whose keys you'll probably never find. But those locked doors will always represent possibility. Or a convincing illusion of possibility, in a game environment. As long as there's another locked door (and there will always be other locked doors), there's the possibility you'll find another key and open up a new area. Maybe you just haven't found it yet. I've heard people say "Every unopenable door should be a scenery door." In my opinion, that takes away half the reason for putting the door there in the first place. Sure, it looks like a door, but everyone knows there's nothing behind it but a blank wall. They know it's not a door door. If it's locked, it not only looks like a door, but also the player has to think of it as a real door, because they never know whether or not they can open it. To those who like to scourge every corner of a map until they're sure there's nothing left to frob, this probably sounds like a nightmare. To me, who likes the illusion of a large world with limitless possibilities, it sounds like just the antidote I need for my dark mod ennui. When I find a key, I think "Ok, this must go to either that one locked door at the Builder Cathedral or that other locked door at the blacksmith shop - all the other doors are scenery." Again, this is a very artificial, immersion-breaking thing to think. But if all those scenery doors are changed to locked doors, getting a key is only half the battle - then I have to figure out which lock it fits, same as in real life. Did I get it off a Builder? A guy in an apron with a welding mask? Did he have a readable on him/near him? If you want to go easy on the player, the key can have a convenient label inscribed on it, visible in the item title like "Cloister". If not, probably still good to have a title that reminds the player where they got it like "key from Builder". This is what mission authors do already, of course - I'm just making the point that a lot of locked doors won't make it impossible to find your way around the mission. If anything, having a only a few real doors among a sea of fake doors is a crutch that allows the player to navigate the mission as an artificial space with a few "interactive nodes" ("I got a new key, so I should go to one of the locked, non-scenery doors"), as opposed to navigating the mission like a real world location, that is, by applying logic and context ("I got a key off a Builder, so I should go to the Cathedral"). In a mission with many locked doors which may or may not be openable, the only sensible thing to do is come up with a strategy to achieve your goal - be a Planner. You may take targets of opportunity along the way, and you may see opportunities you never get to take - who knows whether or not you ever had a chance, just like in real life. No point worrying about the bars of your cage, because you'll never know them when you find them. Of course, unopenable locked doors are just one way of creating the illusion of a limitless world. I'm sure there are other ways. We just have to consider options that may have been ruled out by conventional wisdom. Don't get me wrong, I'm not throwing Explorers under the bus. I actually think the spirit of exploration is better served by uncertainty. Maybe not the spirit of scourging every nook and cranny, though. Sorry Scourgers, I'm throwing you under the bus. It's okay though, basically 100% of all thief and dark mod FMs thus far accommodate scourging. No reason not to have a few Planner-centric missions as well.

"1. This is part of a campaign system, which is in the works now." Sweet! But make sure to allow mappers to mess with the player's gold amount between missions (cap it down to some maximum, bump it up to some minimum, add or subtract a flat amount, etc, presumably with appropriate explanation in the story - some got stolen/had to be spent on X, you sold a non-loot item you picked up in the mission/you extorted someone with information you found out in the mission, etc), so mappers can address the potential difficulty problems Johannes Burock mentions. They shouldn't have to remove loot which they like the placement of, nor cram the mission full of too much loot, in order to calibrate the difficulty for the next mission. Theoretically, you could address this problem by simply allowing the player to change the cost of items in the store from one mission to the next, but in extreme cases, this might seem ridiculous - can the supply of broadheads really be so low they cost $1000 apiece? Certainly, adjusting the item costs is preferable to messing with the loot amount, when it's within reason.

More problems: 1. Overlapping health potions are not cumulative. I had next to no life, and then I used 4 potions in rapid succession, and when they all finished healing me, I still only had half life. The difficulty setting was the lowest one (normal). If I had used each of the four potions after the previous one had finished healing me, the four of them would have been more than enough to restore me to full life. Healing potions should heal the same amount of life regardless of how their healing periods overlap. 2. Lantern light source gets clipped while crawling through vents. When I crouch and crawl through a low tunnel, my lantern light cuts in and out as I bob up and down, I assume because the light source is passing through the ceiling. This could be fixed by lowering the light source coordinates when the player crouches. The light source also sometimes cuts in and out while I'm walking right up against a wall, I assume because the light source is passing through the wall. To fix this, you might need to make the light source a physical object held in place by a spring, so it will get pushed outside the wall rather than pass through. Or you could make the collision shape for the player bigger and put the lantern light source in the middle, so there's no way it can pass through a wall, but of course making the collision shape for the player bigger could adversely affect gameplay in other ways. 3. Items get used if you press "use item" to close a readable. I can press any key to close a readable, but sometimes if I press the "use item" key, closing the readable simultaneously uses an item. For example, I was reading an immobile book, and I pressed "use item", and then the book closed and simultaneously I laid a mine.

That would be cool. In the meantime though, I wonder if there's a way to circumvent the race altogether. I've been thinking about this for a while. In a paper RPG, the DM can draw a crude sketch of a dungeon map or a dragon or a piece of equipment, and the players don't say "Yuck, what terrible graphics." They know the sketch isn't supposed to look real at all - it just prompts them to imagine the real-life version. Quite immersive worlds can be created this way. What if there's an equivalent for video games? It could be something as simple as a post-process effect like the "chalk and charcoal" filter from photoshop, which makes the screen look like a crude drawing. I wonder, as the game progressed, would this continue to be jarring, or remain a mere novelty, or would the player eventually start seeing the game like the DM's sketch, that is, as a springboard for imagination? Would they start imagining how the game world would look "in real life" - more photorealistc than (current) game engines can render?

I think photorealistic graphics are a red herring. Sure, they would be very pretty to look at... and that's about as far as they go. Graphics quality does not make or break atmosphere, in my opinion. Just look at thief - you exist in a world of walking chess pieces, and it's loads more atmospheric than most cutting-edge games today. I'm more excited about advancements that open up new level-design possibilities, like the large, densely-populated worlds of Assassin's Creed. But honestly, I think all you need are the basics in order to create the best possible atmosphere - thief is a perfect example. And judging by how un-atmospheric most AAA games are (in my opinion), better graphics aren't much of a crutch. It's all about level design and interactive storytelling. I mean, you can get utterly drawn into the world of a paper roleplaying game or a book, with no visuals at all. Also, we may already be past the point of diminishing returns, in terms of graphics quality and atmosphere. The last hurdle to suspension of disbelief dwarfs all graphical flaws in terms of its detrimental effect on immersion. Of course, that last hurdle is the fact you're sitting in a chair watching the world on a screen, and not, you know, actually there walking around in it. When I think about incremental improvements to graphics, I imagine someone trying get over a castle wall by standing on the retail box for a graphics card, then standing on a slightly larger graphics-card box, then a slightly larger graphics-card box... Seriously, would the same dark mod mission be hugely more immersive even if the graphics were 100% photorealistic? The only way the player is ever going to get over that wall is if the game engages their imagination and makes them transcend the purely visual experience. Then it's as if they can fly into the air - there are no technical limits to how much they can suspend disbelief using their imagination. Of the different hooks drawing the player into the game world, the visual hook is nowhere near as powerful as when the player meaningfully interacts with the story, in my opinion.

I wonder which idtech4 they're going to open source - the original doom 3 version, the more advanced quake wars version, probably not the even more advanced prey version... hmm.

Hmm, not sure I like the direction of the forum thread linked in that bug. They seem to be talking about adding a special-case solution for glass doors, instead of a general-purpose solution that also works in the examples I mention. Separate visportals and audportals would fix all the cases, including glass doors (audportal only.)

Visual and audio are both controlled by visportals. There should be separate visportals and audportals. This is a problem with the editor, but it affects the player experience as well. Often, sound propagation is odd. Sounds come directly from solid walls. Actually, it comes from the sound source on the other side, but the sound is not routed around the wall because the mapper did not place visportals. The mapper did not place visportals because there was no need to, from a graphical performance standpoint (the original point of visportals.) Consider a short passage with a bend or two, and doors at each end. While you're in the passage, sounds from AI also in the passage will not follow the passage - the sounds will travel through the walls directly to the player. Unless the mapper adds several extra visportals inside the passage. But excessive visportal use can actually decrease performance (right?) - there is no need to place several visportals in the passage, from a graphical performance standpoint. The mapper should be able to place audportals in the passage instead. Also, visportals should not automatically be audportals as well. Consider a large, dark open area, with visportals in "artificial" locations. By "artificial" I mean visportals are not on doors or small openings between larger areas; for example, distance-based visportals could be placed between trees in a forest or between pillars in a large temple, purely to break up the space for graphical performance reasons. This would unintentionally create strange sound directions because sounds would propagate from one visportal to the next, instead of straight from the source to the player as one would expect in a large open area. The mapper should be free to place visportals without worrying about sound propagation. Does this makes sense? Is it plausible from a coding standpoint to have separate visportals and audportals? Perhaps unifying visportals and audportals was meant to simplify the process of map creation, for those not interested in the differences. That's fair enough. Would it be possible to use this solution: Keep the existing visportals exactly the same, that is, affecting both visuals and audio. And then also add 2 new objects to dark radiant (say "simplevisportal" and "simpleaudportal") which affect only visual or audio respectively, for those that want to use them separately.