Is Procedural Animation Worth it?
ฝัง
- เผยแพร่เมื่อ 13 มิ.ย. 2024
- This video is my longest. It gets a bit meta at parts and if you wanted some sort of economic analysis of procedural animation you wont get that here.
My IK Tutorial • Learn Inverse Kinemati...
My Bird Demo benjaminblodgett.itch.io/proc...
Discord / discord
Patreon / benblodgett
Codeer / codeer
GDC talk • Animation Bootcamp: An...
0:00 - Intro
0:25 - Common Misconceptions
3:23 - Disadvantages
5:32 - Scope Creep?
9:03 - Rain World
11:03 - Spore
13:47 - Difference Between Traditional and Procedural
16:23 - Closing Statements - เกม
You know what games would best benefit from this? Parasite games, cosmic horror, biologicla nightmares like The Thing. Those game genres would work amazing with how body parts would just switch bodies and morph or combine in a nightmarish mess. Glorious
I've been handicapped by the lack of ability or resources recently because my 2d rts game is entirely based on this exact approach.
Pretty sure carrion has something like that
@@senorclown9882 Yeah the whole movement in carrion looked and felt procedurally generated, aside from the human characters.
Barotrauma has all three, kinda, the husk infection is a lot like the thing
I think in case of Rainworld, the procedural anims are a big part of the ghosty aesthetic of the world. People in feedbacks praise story etc. , but underneath the hood, it is the aesthetic that makes everything else work. Player feedback is valuable, but take a grain of salt, because people don't really know how to pinpoint what is it EXACTLY why they like a certain game. It might be something like: "It just feels right", but they can't say why it feels right. It would be quite reductionistic to say that "Rainworld was big because it has a great story", because if Rainworld had been an ASCII art game, I doubt we would be talking about Rainworld. Everything matters!
9:17 ok I am just here to say something… Rainworld is not a metroidvania, like sure, it’s open world (for the most part) but it’s still not a Metroidvania. A Metroidvania has you upgrade and unlock stuff throughout the game that allows you to enter new areas and fight new bosses, and Rainworld doesn’t have those factors to make it a metroidvania. But alas, people can agree and disagree with what makes a game a Metroidvania, but I don’t care what you think or say, you cannot convince me this game is a metriodvania
The only metroivania aspect is you upgrading your max karma level
A little... correction? Addendum? Inverse Kinematics is widely studied not only by its application in animation, but also because it's essential in building robots - the kind of robot that sustains our post-industrial society.
I mentioned in the video that there are tons of IK resources partially because of the animation applications. This is because there are lots of resouces from robotics as well. Generally people in robotics dont mention computer graphics in their resouces and the same is true for computer graphics researchers. In the end it's all just math, but I find computer graphics papers are more helpful resources.
I am an amateur enthusiast but in my experience the overlap is pretty minimal. Game IK needs to be realtime and relatively unconstrained. Modern robot IK needs to have collision constraints + other goals and runs orders of magnitude slower (for non-analytical methods). Modern robot IK (think Boston Dynamics, not industrial robot arms) often will involve more than 6DOF whereas a kinematic chain for a hand or foot in a game will probably be like 3-4 DOF (1 for knee and elbow, 2-3 for ball and socket of hips and shoulders)
I may be wrong, but I am not aware of Jacobian methods or optimization methods used for IK in games.
you sound like a nerd
you got a subscriber, nice work
I was blown away when I came to know about procedural animation from the game "Rain world" and felt the urge to learn it no matter what
but seeing this video made me reconsider the huge task I was going to get myself into ,
thanks
I was here after learning of rain world too! I would enjoy a more in depth explanation on the way it was handled in that game but formatted so that I could implement it in godot in 2D but no one seems to do that. I know it's a lot to ask for but if you're not plumb tired of this stuff you should check out rain world and see if you can make a video about it. PLS! *Puppy eyed emoji* XD
I discovered Rain World. I cried.@@dofeila4903
Great video. Clear explanations with good examples. Very informative and entertaining throughout!
A lot of games would benefit from procedural animation, even if only from the side-effects of it. Something as "simple" as the Sims. It's a bit annoying that all the Sims are the same height. But they have to be the same height so the animations will work on all of them.
That was an interesting video. It popped up on a search I did when I was trying to tell someone about the disappointment of Spore's procgen animations being different from what they had shown off previously, and that's kinda historical now. Since I was looking for procedural animation, it brought me to your video and a couple other interesting ones I hadn't seen, and some others I have.
It does seem like your point here is more basic, though. There's a lot of things gamedevs new and old will just try to cram in there, without considering the costs or the benefits. Most video games, especially indie games, don't really need a lot of fancy stuff to be fun and look good and tell whatever story you've got.
My own interest in procgen animation is the whole 'Make Animators Obsolete' ideal. I don't think that'll ever REALLY happen, but it would be nice if we could make a model, rig it, and have it figure out how it could/should move somewhat on its own. Even if it isn't perfect, that would be enough for a lot of folks, and it could output something tweakable probably.
It's also connected to my interest in procgen as a whole. It's always weird when you've got a bunch of copies of the same model for an enemy or whatever, exact same proportions and everything. Then you see a 'humanoid' character with mix-and-match clothes and they look better, even if they have the same proportions. There's also player characters with all sorts of appearance sliders... I would like to see wilder variations on species. If there's a bunch of muscly ogres, I want them to all have different sizes and proportions, and be distinct. I also don't want them all playing the same basic animation like a team of syncronized dancers. Especially since I tend to prefer exaggerated cartoon character types, there's a lot of room for variations that would affect their stride or arm movements. They might not even be symetrical.
Anyways thanks for your contributions to procedural animation, and thanks for reading.
His GDC Talk was AMAZING! YOUR VIDEO IS AN AWESOME COMPLEMENTARY❤
great video! thanks for all the explanations and visuals. this made me think again and again if I really want to get into procedural animations and if my reasons are justified, then I realized as you said at the end, I dont actually need justification. I think its cool if applied properly and I just want to have it on my belt for possible future projects. hope you keep making videos, thanks! ^^
Great video! It was great hearing your thoughts on this topic, keep it up!
Excellent video! Your channel is definitely underrated.
I agree
I agree so much. Gosh, I stumble on so many video "essays" talking about videogame concepts that they know nothing about (or, that they just made up). This video is much more precise and has much more research being the scene. Moreover it has good visuals and explains things clearly. Good job ! I hope you'll get more view in the future.
Sees title, clicks, 23 seconds in, subscribed.
A thing I would like to point out in the game "Rain World" is that I believe that procedural animation was totally necessary for the game, because in this game you basically are a prey animal in a whole new ecosystem new and alien to what the creature is used to, so the animation makes all the creatures feel ALIVE because they basically move however they want, and it can make for scenarios that make the enemies feel like genuine animals!
Like one time, when I was going through an area and then I stumbled upon a MASSIVE tentacular monster which appeared to be blind but not deaf (it is called a "Daddy long legs" in the game), so basically because it was getting towering over the whole room and I wanted to go to the other part of the room, I had to surround him, but here is the catch and why this creature felt alive… as I was trying to hide from it after he spot me below a platform while he was on top of it, this blind octopus knew I was just below him, so instead of going towards me, the idiot just blocked both the exits with it's tentacles and waited for me to just run into it's tentacles.
The thing is that if it wasn't procedural, this thing would've just ran towards me since it knew where I was and I could've escaped, but because of the mix of the AI of the creature and the movement of the tentacles, it makes it feel like a genuine predator
this seems like it was for me ive been obsessed with that gdc talk and love all those proc anim games mentioned and i dont have a math background so ive been just unable and tortured by trying to put this into everything so thanks maybe i'll reassess
Of course rain world is here and I’m glad
Hey man, great video. I'm only dipping my toes into theories for a procedural animation system at the moment, but I'd definitely disagree regarding the differences between traditional and procedural. I believe that can be clearly quantified by the amount of runtime code applied to keyframes while the animation plays, as well as how thoroughly 'baked' the keyframes are, if they even exist at all.
Peak 'traditional' animation would entail as much 'separation of powers' as possible between engine and interpreter: the engine sets the criteria (mesh, rigging, orientation, timing, etc) and then the interpreter animates it by loading a file without making any other changes; effectively, setting the stage and allowing the animation file to play exactly as predefined. I'd liken it to playing a video file in VLC. Everything is baked to a crisp, like a particularly tough cookie, easy to package and deliver.
As soon as the engine starts to apply changes to the animation itself, even if its as simple as layering a different animation over part of the mesh, such as the face or feet, you start venturing out of the 'black' of traditional animation and into shades of gray, which grow lighter as more code is necessary to fully execute the animation, moving towards the 'white' of procedural animation. I'd liken it to playing a video file in VLC with filters and effects applied. Not everything is baked to a crisp, like a softer cookie or chewy brownie, more delicious but messier to package and deliver.
Peak 'procedural' animation would, by my reckoning, *completely ignore* fixed temporal instructions like keyframes in favor of constantly-changing runtime functions that navigate characters through a given environment with as few assumptions as possible. I'm not sure if any of the games you showcased in this video would actually meet that criteria; such code would be horrifically inefficient for interactive gameplay. Something at the absolute pinnacle of 'procedural' animation would probably be difficult to distinguish from a full body simulation, and if it ever tries to employ modern high-poly meshes, it would probably struggle to render one frame every dozen seconds on an average gaming PC. I'd liken it to streaming video to VLC from an AI-generated movie rendering in realtime on local hardware. No baking can even take place, like raw cookie dough, its amorphous and fragile to package and deliver.
Sorry - I kinda got carried away with the metaphors. Now I'm hungry lol
But yeah, I definitely agree with the brunt of your sentiment - most well-polished 3D games of the past decade would average out to some darker shade of gray on this gradient between traditional and procedural, with some lighter than others. I also believe that for a well-polished, and perhaps more importantly, *well-optimized* modern 3D game, you ideally want a mixture of both, and you want to pick whichever is most well-suited for the task at hand. Nobody would want a game's engine trying to run fully fleshed-out simulations on every movement of a 3D mesh in its environment - that would be a nightmare for a developer to code and a nightmare for a player to operate. So even the most ambitious procedural animation games will have to make plenty of compromises for the sake of efficiency and playability. Those that don't will just end up becoming memes - like QWOP, but without the self-awareness. I'd argue that this is why so many games that attempt to leverage highly procedural runtimes end up with relatively low polycounts and minimalist stylizing - its not always an aesthetic choice. More often than we realize, it's probably a corner they've developed themselves into, wittingly or not.
One of the projects I'm currently theory-crafting starts with a traditional keyframe-based approach, with optimized/packaged files and an engine with an interpreter that reads them, but with a procedural flair where any value of any coordinate in a given keyframe can optionally include variables, which default to zero, but can be updated dynamically at runtime, ideally with data from both the skeleton and mesh itself. For example, consider a simple high-five between two human characters: the animation file wouldn't know or care about skeletal factors like height or appendage length, or mesh morph factors like body weight, but the animator will know that this can differ, so they can include variables in keyframes for parts like the collar, shoulder, forearm, and hand. (Possibly many more, in case you want characters to range from Michael Jordan to Peter Dinklage) Finally, the engine, which has all the data it needs about the mesh and skeleton, calculates the necessary values for these variables - ideally long before the animation is actually loaded - and pipes those values to the animation interpreter, which factors it into its calculations. This should hopefully maintain the optimized workflow and streamlined execution of traditional animation, but with a healthy layer of dynamics when needed, akin to procedural systems.
Probably the main thing preventing me from better understanding the distinction between procedural and traditional is Ive never written a system for rigging meshes and playing animations or any of those sorts of low level details for traditional animation. Its does seem to boil down both of these types of animation techniques exist on a spectrum, but it's hard for me to conceptualize a better distinction than "you know it when you see it" with my limmited experience. Thanks for the comment!
Thank you for video!
Great video! I need to take some time to try to imlement some procedural animation myself
Watched this very shortly after watching that GDC video mentioned. Good stuff
hey man this was an awesome video and super inspiring. I want to make the best possible game I can make, I am making a piece of art that shows my love for detail, a game that i dreamed of :D I feel like procedural animation is the future, and once tuned right can offer amazing things. But like you said the time trade off is important. I'm super driven to make this great, and I've learned everything else so far, so this won't stop me.
I really appreciate your video. It was truthful, balanced, FUNNY, and high quality. thanks dude, keep it up!
I still think Rain World's procedural animation is very well deserved and the game wouldn't work otherwise. It is fine that people don't mention it in the reviews as the procedural animation still makes the world feel alive as it should, and people definitely see the world as alive.
dude I love spore too, I always imagined a game like spore but mixed with the sims, where all interactions are welll tuned procedural animations. Thats my goal! Man this is such a good video
I am NOT a game dev YET. but I have a whole concept of a game and "physics" animation, animation that bumps into stuff and so on is essential. the game will be very vertical with lots of terrain, and I want the main way for the player to interact with the world being moving its affected by physics body. its an alien ecosystem, the whole concept was inspired by rain world, I already had a spec evo project before, but rain world inspired me to create a game concept, and I want the world to feel organic and dynamic, ever changing, behaviour = animation, and the interactions creating the a.i responses.
adding to that, I'm really more of searching for something that can be affected by collisions and such, of course its better to fake it than to really make it (cause the inverse kinematics interact with the terrain "collide") so its practically the same interaction as if I attached a bunch of sprites together and made it move using those, but it would take a lot more processing power for the same thing, wouldn't it?
In other words, how can I make the animation more interactive? for example, allowing a creature to be hit on the legs and loose its function?
Hi ty for the video! I want to make a monster taming game where you fight as the monster itself and can move freely in battle the battle not turn based you have all kinds of skills and you can dodge and attack and it's going to be online do you think it's going to be a good idea to do it with procedural Animation?(i'm really bad at making art and animation )
I didn’t realize how procedural animation was defined until you said gang beasts.
I wouldn’t deter anyone from learning this method because it looks so cool.
Also this has been done since the flash games era believe it or not.
4:18 what is theese worms or something like that i want to see it
This is a hidden gem! Needs to be seen by much more people!
I totally understand what you mean by not loving making games but loving making challenging system. It is like a puzzle game and there is no point in polishing the game since the puzzle is already solved
i love this video
Great video! The scope creep segment was particularly useful.
What is the game shown at 4:16 ?
Thanks! Glad you enjoyed it.
It's from this 4 year old reddit post www.reddit.com/r/proceduralgeneration/comments/6gwc0m/physics_based_creatures_with_exchangeable_body/. This guy inspired Rujik to make his proc ani stuff, but other than that his demo doesn't seem to have amounted to anything and his reddit account is inactive.
@@BenjaminBlodgettDev Thanks! This is super helpful.
I'm working on a top down game where the player has a Ball and Chain as a tail, so twirling around in a circle swings the ball and chain, and the player can swap out the Ball at the end of the chain with parts they knock off of enemies, so when I saw the clip I immediately thought it would be useful.
@@ChrisSmith-bb1dq That sounds really cool! If you make any progress I would be interested in seeing what you get done. You ought to join my discord and post some screenshots there
We FINALLY HAVE UE5 MERGING BOTH ❤❤❤❤ THANKS FOR YOUR VIDEO, IT'S AMAAAAAAAZING ❤
Do you think an FPS or TPS should use it for ADS, leaning, and other stuff like that or is best to stick to traditional animation?
A weird example of procedural animation is Source of Madness. I didn't like the gameplay so much (enemies being such a procedural mess actually made them feel same-y). HOWEVER the concept and idea behind generating infinite enemies is very cool, and I wish more games would explore that concept. I will for sure be following the devs in case they throw out a new game/update with improvements.
It's a very cool game that I think is close to a masterpiece. A few game design flaws and dev inexperience are holding it back (they are experienced, but it's ambitious so it takes more).
Came here because Rain World though, 10/10 good DLC good game.
Fascinating! What is the game shown at 1:21?
Hey, Great job and editing is fun. Thank you for entertaining me and educating me.
you read my mind.
0:31 Oh hey I clicked on a random YT video and is that my own comment I see??
Excellent video, you mention that procedural animation looks worse and equally people won't notice which does then become a mute point? So then it really comes down to effort, to me, I'm not a game Dev, my game is a tech demo for me, I have no interest in sharing with others so that puts my choices into an unusual, it's just for me state, I would choose to try to emulate Davids work, because it absolutely blew my mind
My Player have multiple part that have sepert animation but my player is robot and don't look weird the part I is the following.
1. body/Head
2. The lower half (legs)
3. The hands.
The reason I have done that is because I'm working on fps game and I want the Player to make extra animetion for every little thing.
And because my player is robot I have more room for error's.
To be honest my Player mixed between skeleton animation.
The old type of animation that some part animate indepeding.
And I don't know this consited as animetion but some parts move with Player action like for example wean player move the mouse in Y the move the rooted head/body up or down alone with the head.
The alose many other parts I have plans to add animation like camera wean close to explain ECT.
If you ask that's proof my have plan to advance animation system even if do that in the order to save time
Because make animetion with skeleton animation is very hard for My 😅.
Especially the Player is human.
Ther I reason that I ended make robot 🤖 originally I have plan to make hummun as player wean cume to animet the character I have realized how hard the are,
And I disait that I will make my game play to feeting animation instead and the esyer way to do that is if the Player is robot 🤖.
I don't know how this part aprch called but it's interesting way for animation.
The only reasons I would use procedural animation is because I'm terrible at animating things by hand and because animation is extremely time consuming and tedious. I don't want to spend an hour or two making a ten second animation.
EDIT: Also, when you are talking about inverse kinematics, I am guessing you are not referring to rigging with a bone structure in a 3D application at all? Are you talking about programming the rigging as well as the animation?
How many times do you want to repeat yourself?
Benjamin: YES.
sorry. this was one of my older videos where I experimented with having no script. not entirely happy with the result
1:48 where is that from???
i think that justfiying why you're using something in your game is only really necessary if you want to finish the game in an effecient manner ; and we already know thats gonna fail so might as well not do that.
there is no efficiency, only trial and error, so might as well learn it and allow yourself to be sidetracked, who cares? just make sure to document whatever idea you left behind to come back to it once your procedural fun is over.
Some good gamedev points for the fact that you don't like game design ;d I love the meticulous work on the presentation. Very enjoyable watch.
To my mind, if you have alot of dynamic content, it's probably gonna be the same amount of work as if you had done it traditionally.
I would want to watch this video even if I was not interested in this topic ..... You' just made me laugh in 20 seconds
I understand the difficulty of conveying the correct point through text sometimes. So, fresh start, I'll try and reply piece by piece:
"I'm sorry if my reply came off as belittling."
Thank you. It did come off that way, but the apology cleared it up.
Regarding the confusion: I was responding to your intro, to the joke/meme in the first 20 seconds where you depict a commenter responding "but isn't procedural animation just IK?"
Your hypothetical viewer reminded me of a related topic. I wasn't replying directly to something you stated in your video, or attempting to correct you in any way. I was literally just bringing up an adjacent topic: That the broad over-use of the term IK causes newer developers to be unnecessarily confused about what it is.
To summarize without a billion examples: IK is "whatever you can do with IK" by common language, even often in IK specific tutorials. We refer to types of keyframing, rigging and animation modification all by the same umbrella term. So new developers end up considering it a black box. Because if they watch a thousand talks about Inverse kinematics, they may never realize that the term just means "inverse of forward kinematics: Translating changes from child to parent, instead of parent to child." And, at the risk of repeating myself, it's because IK is "whatever you can do with it." It's motion warping, it's procedural animation, it's keyframing a hand yet animating a whole arm, it's slope matching, etc. Because people CALL those things Inverse Kinematics all the time, without stopping to explain the core principle. Thus, new animators have no idea what IK even is for far longer than is needed.
I repeat: This is an adjacent topic. I wasn't criticizing you or even responding to you directly at all: Your intro made me think of a similar, but not identical topic. And since TH-cam likes comments, I felt it was best to leave the engagement rather than internalize the thought.
yeah I definitely agree! I think a similar case is raytracing. People get so caught up with the idea of physically based raytraced rendering and "rtx" that they forget that raytracing is just returning the time of intersection of a projected ray. Thanks for clarifying your response!
Equally, cool animation is using Verlet integration.
So as a Blender user you are saying that I should avoid procedural animation, but use it where feet and hands impact with objects. Inverse Kinematics is not something I will never avoid using in Blender.
Procedural or not its just tools, you can view trad animations like bunch of curves that dictates movement and use them, like in UE there is timelines where you can specify how variable can change over time.
Great video! it really goes in depth of procedural animation.
what is the game shown at 3:02
Thanks! its Exanima
2:53 well, to be fair, no one really needs art..... thats why is so beaultifull.
.
.
.
.
IMMA FUCKUING USE PROCEDURAL ANIMATION ON MY DUMB GAME OK????
wtf how dare you kill beloved character snail from franklin
...however you do mention examina which is awesome. You are safe, for now.
3:09 Wild malapropism spotted!
Really good video, liked it a lot. I enjoyed the chaotic feel of it, it kept me on the video more lol. (Please don't take this badly, but rain world isn't a metroidvania) ^^
I would prefer 'keyframe animation' term instead of 'traditional'. That also makes a better definition for the animation types. Everything that uses some sort of code to animate would be procedural in my opinion.
David Rosen's technique revolves around keyframes and that seems to me obviously procedural. Also even during runtime non procedural animation requires code to animate. So neither of your criteria help disambiguate between what is procedural
0:20 Oh i see you're too man of culture ;)
I think AI-assisted traditional animation has much more potential than procedural animation (see Cascadeur).
Jesus you need to tone down those topic introduction song in this video. Otherwise i liked it very much and agree on many things.
A space game with spherical worlds could benefit from procedural animation.
Exanima is awesome!
Only the strong can really do it.
All those AAA+ games announcements... Everything turns into shit as soon as anything starts to move. Including the camera tied 1:1 to mouse input!
It feels utterly bizarre to hear any form of 3D cgi being described as traditional animation. The only form of 3D animation that most would consider traditional is stop motion, lol.
Rain world good
Lizor spin
@@robertchappell8086 based
My god.. im not making a game. Im not making a game im not making a game!
I
Everything is about game, power fantasy. I need to make a living individual piece. Always with the games, i have to reverse engineer everything i hear
Every book ive taken out talks about a game.
What are you trying to tell, what will I get when I watch the video? Just pointless talk. I watched 15 minutes and got bored.
Yeah. I did not write a script for this video so it was very ranty. I apologize and hope to improve this in my newer videos. Thank you for watching though!
Not enough subway surfers or minecraft parkour, I guess.