I'm definitely keeping an eye on this; I'll have you guys do all the work in figuring this stuff out and, and then swoop in and snatch it up for use in my stuff! Hahahaha!
Why not just have two textures per character? One lightly shaded and one heavily shaded, then just lerp between them depending on the light angle.
You'd get the popping anyways because the animation is on 2s/3s, and if you wanna exaggerate the popping then you can have the lerp only do a hard switch between 3 modes (0%, 50%, 100%).
This pretty awesome deconstruction you all have done. Im almost certain it's a combination of the normal technique and using a dedicated light and shadow texture, like Bigjohn said.
I tried something similar to this a while back for a project with relative success, no where near this amazing but the concept is sound I think.
I'm not sure if this is what's used, but i'm pretty sure I can get nearly the exact same results using a faked BRDF: Just use a different texture sampler per material.
I'm not sure how to record it. It looks decent as the light moves. Just gotta make sure that the shadowed version looks like an extension of the unshadowed one (unlike mine). Cause it looks like the shadowed version "grows" out of the unshadowed one. So it has to make sense.
It's a very simple network, should take like 2mins to build in UDK.
I'm sure if someone used it on a proper mesh, with an actual normal-map, and spent more than a minute on the diffuse textures, they could get a decent result.
I am absolutely excited to see a breakdown of how they did this. I also love the kind of discussion that sort of innovation creates. Here's to hoping this invigorates the industry a bit and we see more stuff like that.
The next iteration of Street Fighter should use the same method with the SF3 style sprites. My brain would literally explode.
I THINK I saw another article that talked about this particular thing from Arc System, but it was in Japanese and the Google translate was spewing gibberish. I'll try to find it once I get a minute Troy. Unless I imagined it and my memory is a lie, in which case I'll be back with nothing.
I'm bringing this back from the dead because there has been a location test (ie a public beta at an arcade) in japan and videos, albeit shaky, have surfaced. Very much worth a look as it further shows just how close to the 2D original they've managed to get:
There was a video done at Ubisoft where they showed off the tools they used for the new Rayman game, which are pretty awesome. And at around 14:30, they explain how they used 3D for some bosses and how they made it look 2D:
[ame="http://www.youtube.com/watch?v=y-chi097uV4"]How Rayman Legends Is Made! - YouTube[/ame]
I know I might sound ridiculous but... maybe just maybe Guilty Gear Xrd is in fact 2d... and they just swap to 3d when its supposed to rotate at the start round scene. I know how it sounds but often the simplest most obvious solution is the best... since it is way more simple and effective. Why would they make 3d look like 2d and move like 2d and then slow frame rate to give it a true 2d feel to it... maybe they only failed to make it look 2d at the scene we all noticed it was 3d...
I mean its just like faking a video tape by drawing certain 2d elements on it to the point they look amazingly 3d realistic... with amazing 3d software siting next to you.
I know I might sound ridiculous but... maybe just maybe Guilty Gear Xrd is in fact 2d... and they just swap to 3d when its supposed to rotate at the start round scene.
Well, for one they've confirmed it's 3D, but here's why it makes sense:
-Arcsys games have always used 3D models, just indirectly until now. They create all characters in 3D, animate and render them and then have pixel artists go over the result (similar to what KOF 12 and 13 do). Throughout the games they've perfected the process to a point where P4A characters required very few touch ups.
-Swapping would require having whole 2D and 3D sets for all the movements since it can swap at unpredictable times (specifically, any KO, dust attack, supers and some other moves), not just the start of a fight. Since as pointed out earlier the 2D set is a render of the 3D set it makes sense to just remove the distinction. They look the same, so why waste the memory on 2D? They had nothing to gain from it.
-Polygonal characters bring several huge advantages: they're resolution-independent, take up less memory and alternate character appearances can be created without having to reanimate them. Not only this makes it easy for them to create bonus or DLC costumes, they even make use of the feature in normal game flow: Sol Badguy's dragon overdrive, which used to just be his sprite blinking during a power up, can now be a wholly new physical appearance.
Maybe I'm wrong id be but, and i think i got it:\, as somebody said before they are relying on 2d techniques heavily and i think the lights and shadows are faked , the reason of why the textures look so similar to the 2d sprites is because the light and the shadows parts are all embossed in one diffuse texture, there are no lights affecting the model, apparently the model uses multiple textures and depending on the pose and and keyframe of the character the engine changes to an specific texture for the pose and look correctly shadowed,so is like a 2d sprite sequence sheet, obviously here is not necessarily to paint every single frame with the correct shadow and light if the character is not moving that much unless the model is moving to a very different angle.
Again i may be wrong , i know it sounds like a ton of work but hey drawing sprites in the correct pose,painting the right lighting and all that stuff is a pain in the ass too.
I was lucky enough to be in Tokyo during this second location test and it's beautiful seeing and playing it in person. Definitely almost entirely 3D, and even the bits that wouldn't really need to be (like the face close-ups during that finger poke move) are at the very least painted over an underlying 3D model to introduce subtle perspective shifts.
I seriously don't think that they would waste a 3D engine on a game that plays entirely in 2D and has 2D characters. Even SF4 had 3D characters in a 2D camera shot. I think the game is 2D, they just rendered a few 3D scenes to match some of the action. There would really be no point in making 3D that looks totally 2D. Even cel-shading still looks 3D as far as viewer perception is concerned.
I seriously don't think that they would waste a 3D engine on a game that plays entirely in 2D and has 2D characters. Even SF4 had 3D characters in a 2D camera shot. I think the game is 2D, they just rendered a few 3D scenes to match some of the action. There would really be no point in making 3D that looks totally 2D. Even cel-shading still looks 3D as far as viewer perception is concerned.
There is no point? What about flexibility?
"oh that perspective doesn't work really well, i think might have to change it a bit. Okay lets do this, redraw every pose from another angle!"
it makes perfect sense when you look at it from a production point of view. Of course the initial cost is higher, developing the tech. But once it is there is is much much cheaper to produce and change things in 3d.
"oh that perspective doesn't work really well, i think might have to change it a bit. Okay lets do this, redraw every pose from another angle!"
it makes perfect sense when you look at it from a production point of view. Of course the initial cost is higher, developing the tech. But once it is there is is much much cheaper to produce and change things in 3d.
I get the production value aspect of going 3D instead of 2D, but don't you think if it was actually in 3D that there would be more utilization of 3D perspective? I find it odd that they would create an entire 3D game design and only utilize it for a cinematic effect here and there and not even during gameplay. There is really nothing in the demonstration that is telling me that this is a real-time 3D engine.
Furthermore, why go through the trouble of excluding frames just to make the audience/player think that it may be 2D? For the artistic value? Well, I guess it works. I'm impressed, just baffled at the same time. If it were me I'd just give it the smoothness you normally don't get with 2D and include all the animation frames, it'd look cool.
Games like Sonic Rush operate entirely in 2d, look like they are rendered in 2d, but are actually rendered in 3d. This is because modern 3d art pipelines are much more flexible.
This series has traditionally been 2d and it's a hallmark of the game's artistic direction that things are drawn in a certain way. They've put a lot of effort into ensuring their newest iteration looks and feels like the predecessors to keep their existing fans satisfied and to remain consistent and retain their core values. Their pipeline however lets them do a lot more, more quickly than they could do with previous games. Combo move animation doesn't look right? Don't redraw 20 frames of animation, tweak the skeletal animation. Character design needs updating? Don't modify EVERY frame, update the mesh.
Timidy, there's no doubt that it's a 3D engine: it's UE, as prominently displayed in the trailers. The 3D is actually slightly visible during gameplay even beyond the KO animations (where the camera orbits around the finishing move) as the characters' perspective changes as expected as they move from left to right of the screen. Also you're going about it the wrong way, they aren't excluding frames, as that's how they animated in their previous games anyway (Blazblue and P4A already used 3D models for sprite production, they've just removed the middle man, so to speak), they're merely choosing to animated without tweening, which is a valid, if highly unusual, choice.
There are multiple advantages to this approach. First, the problem in their previous attempts is they were lacking in sprite memory - for example in one of the blazblue revisions they had to drop frames in one of Iron Tager's (a large character) attacks because they couldn't add new moves otherwise. Though arguably their sprite memory management was terrible (they stored the sprites raw and untiled, whereas Skullgirls for example uses excellent compression tech to store twice as many sprites at four times the resolution for six characters instead of two) going 3D allows for skeletal animation, which is highly more compact by magnitudes. Of course the tons of memory in next-gen consoles would have room for all those sprites bu at the location test the present devs said the current version could run on a PS3, which is rather impressive.
This also means their graphics are now all vector-based, and thus resolution-independant. Want to do a 4K version? You literally only need to change the resolution of the display. Their characters are even very detailed already, with small intricate touches in their clothing you only see up close.
There are added benefits to skeletal animation. Pose mirroring is one: the game is able to selectively invert the skeletal poses automatically to account for the left/right facing graphics, and as such the clothes and other details do not flip when the character faces the other way (unless they're meant to, like Slayer's living cape). Using sprites you'd have needed to redo all the sprites or use complex palette effects to fake asymmetry.
Costume alterations are another one: character graphics can be changed without altering the animation data, only the models. It's highly likely we'll see that illustrated with DLC costumes but it's already visible in the game via in-fight alterations. The two best known ones right now are Ky's (the blonde swordsman) ponytail, which can done undone if the character suffers a blow to the head, leaving him to finish the fight with his hair flowing free (you'd have required alternate hair sprites for that before) and Sol's (the red swordsman) Dragon Install, which is a buffing special move: in the previous versions it used to be a mere palette effect when it was on, but now thanks to skeletal animation it switches to an entirely different, spectacular burning demon-like model that still has all of Sol's moveset since it uses the exact same animation data. With sprites you'd have needed an entirely different sprite set.
There's also one additional parameter that's decidedly down to earth: that dates back from a blazblue interview but they've confessed not having any traditional 2D animator on the team - thus why BB and P4A use 3D models in preproduction. Then then outsource the 2D part to China. Given that, it makes even more sense to remove the final pixel art phase in their pipeline since their team makeup is decidedly 3D-oriented.
If they're doing away with traditional hand-drawn 2D animation in games how long until the Simpsons and Family Guy are animated in 3D? Or at least vector-based. The one thing I don't like about that style of animation is it comes off too one-dimensional sometimes, lazy ; like the characters are only able to face one direction and can't, do a Micheal Jackson spin dance, for instance. With frame by frame hand-drawing there are no limits, but it's very tedious and time-consuming of course.
If they're doing away with traditional hand-drawn 2D animation in games how long until the Simpsons and Family Guy are animated in 3D?
The Simpson's did away with traditional hand drawn animation over ten years ago and moved to digital ink. When they'll make the switch to a fully digital or 3d pipeline is anyone's guess though.
Or at least vector-based. The one thing I don't like about that style of animation is it comes off too one-dimensional sometimes, lazy ; like the characters are only able to face one direction and can't, do a Micheal Jackson spin dance, for instance. With frame by frame hand-drawing there are no limits, but it's very tedious and time-consuming of course.
Your last sentence says it all: what you're speaking about is a limitation of automatic tweening, not of vector-based 2D animation. Doing it frame by frame is exactly the same tedium whether on paper, bitmap or vectors. So the vector-based method doesn't have extra limitations; rather, it offers a tweening function that people erroneously hope will allow them to draw less frames for the same results.
Incidentally, that's why some company will use 3D models or 3D-based tweening (like Disney's Meander, used for the Paperman short) instead: the inbetweening for those kinds of moves is of course gonna be right, though it's arguable whether it saves work since you have to model and rig a character (though a very simple one in the case of Meander), for a company like Arcsys the goal is of course to transform the task from 2D animation, which they can't do in-house, to 3D animation, which they can. And of course, the long-term advantages mentioned in earlier posts do make sense for a game.
Looks like they draw the faces over the top imo. Atleast it appears so more in some scenes more than others. I dunno if that wireframe shot is anything to go by since it could have been fairly dense to show wires for the face which could be a reason to draw over atleast that shot even if the rest 'may' have been full 3d. Either way that's pretty damn cool
Every other month in 3D World they go over how various Japanese studios mix 3D with 2D for their 2D animations (or their games). Later on I can scan a few pages then provide some translations for key bits.
Every other month in 3D World they go over how various Japanese studios mix 3D with 2D for their 2D animations (or their games). Later on I can scan a few pages then provide some translations for key bits.
Replies
You'd get the popping anyways because the animation is on 2s/3s, and if you wanna exaggerate the popping then you can have the lerp only do a hard switch between 3 modes (0%, 50%, 100%).
I tried something similar to this a while back for a project with relative success, no where near this amazing but the concept is sound I think.
I'm not sure if this is what's used, but i'm pretty sure I can get nearly the exact same results using a faked BRDF: Just use a different texture sampler per material.
http://www.mentalwarp.com/~brice/brdf.php
Whipped this out real quick in UDK. I think if you combined this with the normal-map lighting thing from a couple of pages back it'd work pretty well.
All it's doing is lerping between those two textures based on where the light is.
It's a very simple network, should take like 2mins to build in UDK.
I'm sure if someone used it on a proper mesh, with an actual normal-map, and spent more than a minute on the diffuse textures, they could get a decent result.
I ended up going with an If node instead of a Lerp. It let me have a sharper transition.
The next iteration of Street Fighter should use the same method with the SF3 style sprites. My brain would literally explode.
+1 for Megaman
Fast forward 17 sec mark for beginning of tank shodown.
[ame="
Fast forward 42 sec mark where turret pov starts to move.
[ame="
It seems just very texture heavy and without much lighting.
Is this the article? http://www.arcsystemworks.jp/official/company/tech.html
[ame="
I thought for sure seeing more characters would eventually show us where it falls apart, but this is so amazing i can't even handle it.
[ame="
Lovers in a Dangerous Spacetime is pulling off pretty good 3D posing as 2D in Unity. http://www.asteroidbase.com/devlog/4-two-and-a-half-ds/
[ame="
I mean its just like faking a video tape by drawing certain 2d elements on it to the point they look amazingly 3d realistic... with amazing 3d software siting next to you.
Well, for one they've confirmed it's 3D, but here's why it makes sense:
-Arcsys games have always used 3D models, just indirectly until now. They create all characters in 3D, animate and render them and then have pixel artists go over the result (similar to what KOF 12 and 13 do). Throughout the games they've perfected the process to a point where P4A characters required very few touch ups.
-Swapping would require having whole 2D and 3D sets for all the movements since it can swap at unpredictable times (specifically, any KO, dust attack, supers and some other moves), not just the start of a fight. Since as pointed out earlier the 2D set is a render of the 3D set it makes sense to just remove the distinction. They look the same, so why waste the memory on 2D? They had nothing to gain from it.
-Polygonal characters bring several huge advantages: they're resolution-independent, take up less memory and alternate character appearances can be created without having to reanimate them. Not only this makes it easy for them to create bonus or DLC costumes, they even make use of the feature in normal game flow: Sol Badguy's dragon overdrive, which used to just be his sprite blinking during a power up, can now be a wholly new physical appearance.
[ame="
Again i may be wrong , i know it sounds like a ton of work but hey drawing sprites in the correct pose,painting the right lighting and all that stuff is a pain in the ass too.
So what do you think guys?:)
first 2 minutes are character intro's and Faust's special moves(funny)
the rest is game play stuff
Cam corder. I can't wait to see Jam!
[ame="
Thanks Bigjohn for showing me how to get the youtube clip to embed
Many videos there, if they haven't been taken down by the time you read this: http://www.youtube.com/channel/UCy6pX5KDWZAtefrabv-etsg/videos
There is no point? What about flexibility?
"oh that perspective doesn't work really well, i think might have to change it a bit. Okay lets do this, redraw every pose from another angle!"
it makes perfect sense when you look at it from a production point of view. Of course the initial cost is higher, developing the tech. But once it is there is is much much cheaper to produce and change things in 3d.
I get the production value aspect of going 3D instead of 2D, but don't you think if it was actually in 3D that there would be more utilization of 3D perspective? I find it odd that they would create an entire 3D game design and only utilize it for a cinematic effect here and there and not even during gameplay. There is really nothing in the demonstration that is telling me that this is a real-time 3D engine.
Furthermore, why go through the trouble of excluding frames just to make the audience/player think that it may be 2D? For the artistic value? Well, I guess it works. I'm impressed, just baffled at the same time. If it were me I'd just give it the smoothness you normally don't get with 2D and include all the animation frames, it'd look cool.
This series has traditionally been 2d and it's a hallmark of the game's artistic direction that things are drawn in a certain way. They've put a lot of effort into ensuring their newest iteration looks and feels like the predecessors to keep their existing fans satisfied and to remain consistent and retain their core values. Their pipeline however lets them do a lot more, more quickly than they could do with previous games. Combo move animation doesn't look right? Don't redraw 20 frames of animation, tweak the skeletal animation. Character design needs updating? Don't modify EVERY frame, update the mesh.
Hehe.
There are multiple advantages to this approach. First, the problem in their previous attempts is they were lacking in sprite memory - for example in one of the blazblue revisions they had to drop frames in one of Iron Tager's (a large character) attacks because they couldn't add new moves otherwise. Though arguably their sprite memory management was terrible (they stored the sprites raw and untiled, whereas Skullgirls for example uses excellent compression tech to store twice as many sprites at four times the resolution for six characters instead of two) going 3D allows for skeletal animation, which is highly more compact by magnitudes. Of course the tons of memory in next-gen consoles would have room for all those sprites bu at the location test the present devs said the current version could run on a PS3, which is rather impressive.
This also means their graphics are now all vector-based, and thus resolution-independant. Want to do a 4K version? You literally only need to change the resolution of the display. Their characters are even very detailed already, with small intricate touches in their clothing you only see up close.
There are added benefits to skeletal animation. Pose mirroring is one: the game is able to selectively invert the skeletal poses automatically to account for the left/right facing graphics, and as such the clothes and other details do not flip when the character faces the other way (unless they're meant to, like Slayer's living cape). Using sprites you'd have needed to redo all the sprites or use complex palette effects to fake asymmetry.
Costume alterations are another one: character graphics can be changed without altering the animation data, only the models. It's highly likely we'll see that illustrated with DLC costumes but it's already visible in the game via in-fight alterations. The two best known ones right now are Ky's (the blonde swordsman) ponytail, which can done undone if the character suffers a blow to the head, leaving him to finish the fight with his hair flowing free (you'd have required alternate hair sprites for that before) and Sol's (the red swordsman) Dragon Install, which is a buffing special move: in the previous versions it used to be a mere palette effect when it was on, but now thanks to skeletal animation it switches to an entirely different, spectacular burning demon-like model that still has all of Sol's moveset since it uses the exact same animation data. With sprites you'd have needed an entirely different sprite set.
The Simpson's did away with traditional hand drawn animation over ten years ago and moved to digital ink. When they'll make the switch to a fully digital or 3d pipeline is anyone's guess though.
Your last sentence says it all: what you're speaking about is a limitation of automatic tweening, not of vector-based 2D animation. Doing it frame by frame is exactly the same tedium whether on paper, bitmap or vectors. So the vector-based method doesn't have extra limitations; rather, it offers a tweening function that people erroneously hope will allow them to draw less frames for the same results.
Incidentally, that's why some company will use 3D models or 3D-based tweening (like Disney's Meander, used for the Paperman short) instead: the inbetweening for those kinds of moves is of course gonna be right, though it's arguable whether it saves work since you have to model and rig a character (though a very simple one in the case of Meander), for a company like Arcsys the goal is of course to transform the task from 2D animation, which they can't do in-house, to 3D animation, which they can. And of course, the long-term advantages mentioned in earlier posts do make sense for a game.
[ame="
http://www.crunchyroll.com/anime-news/2013/12/03/video-kamikaze-douga-produces-high-quality-anime-pv-future-photon-harima-sacla
Looks like they draw the faces over the top imo. Atleast it appears so more in some scenes more than others. I dunno if that wireframe shot is anything to go by since it could have been fairly dense to show wires for the face which could be a reason to draw over atleast that shot even if the rest 'may' have been full 3d. Either way that's pretty damn cool
Please do, Thank you
That's rendered with Lightwave3d, not realtime.