You’ve got to admit there is a certain magic about the feel and look of classic hand-drawn animation especially when it’s done well. I’d prefer well done 2D animation over 3D any day of the week though don’t get me wrong, I love them both!
We’ve all seen 3D rendered as some sort of 2.5D abomination that usually looks abysmal unless a stylistic touch is severely used resulting a fairly good result.
We’ve seen the industry gradually move away from 2D animation and to 3D, and for good reasons too. 3D allows greater control, takes a bit less time, can look amazing, and can be integrated seamlessly with live action in the forum of special effects. It also has a larger base of people who can actually do it or at least a division in, for example modeling.
What if the two could be merged in a perfect fashion? Done as 3D, made to look 2D. You’d have all the benefits that 3D entails while being able to capture that magical 2D feeling.
I find the chiefest among issues when it comes to previous attempts at this to be the following…
Interpolation - traditional animation has a stylistic, choppy feel, usually 3D rendered as 2D is animated as if it was 3D, BIG mistake! That makes it look really weird! You’d have to animate it with a 2D mentality.
Complex camera movement - 2D animation “cameras” you could say, usually only do simple 2D movements, any movement that involves depth tends to be as simple as possible.
Complex lighting - usually lighting and shadows in 2D are fairly solid and only enough to define shapes in a basic manner. complex lighting like you’d normally find in 3D causes it to look weird as 2D. Lighting a mesh with a few simple lights, trying to keep all shading and highlights simple is ideal.
Materials that are too complex - 2D usually involves basic colors, textures, and gradients, almost the opposite of what you normally find in 3D.
I rest my case. I’m going to look into this and attempt it.
Blender has cell shaders that won’t give you gradient shadows. You can also simplify camera movement. One thing I’ve noticed about camera movement in 2D animation, is that it frequently takes place over a still flat matt painting. To emulate this in 3D, you’d need to make a render of the scene, then pan the camera over the render, rather than the scene.
Good luck with your endeavor. Looking forward to seeing your results.
If a company with the collected experience and wisdom of Disney has trouble pulling this off, I don’t like your chances.
Any CG animator worth their salt with enough time to finesse their work is in constant battle against lifeless interpolation though. The respective processes and approaches aren’t hugely different up to the point where in-betweening takes place - in 3D/CG, the computer does the in-betweening based on keys, extremes, breakdowns and tweaks set by the animators; in traditional cel animation, it’s done by lowly assistants based on keys, extremes and breakdowns drawn by the animators.
It’s common for cel animation to work on “twos” during slower actions that don’t need as much resolution in “time”. There’s nothing stopping CG from using twos - David OReilly uses them a lot in “The External World” and they look great.
But where 3D and full-motion cel really part ways is the cel animator’s freedom to directly create the shapes we see on the screen. 3D CG creates the shapes at more of a distance by projecting shapes and colours onto the picture plane by using virtual objects, light sources, texture maps, etc. You can be reasonably expressive with CG characters if you’ve got a detailed enough rig and you’re willing to put the time in, but you can be much more expressive with a series of good drawings - especially if you’ve already learnt, memorised and forgotten the Twelve Principles Of Animation.
If you want to do 2D animation though, why not just learn to animate in 2D? 3D’s a bigger pain in the arse: it takes many more people to make it, it takes more time, it costs more money, everyone complains it’s not as hand-crafted as 2D…
Not necessarily the case - this is where the “Shift” parameter in Blender’s camera comes in. Depending on how the scene is built and the length of the lens, you can animate the “Shift” parameter on the camera to do a fake pan. However if you shift too much in any direction, the perspective starts to distort.
The old theatrical shorts did some incredible perspective tricks with their backgrounds on shots with a lot of camera motion to do things like changing the camera POV from looking up from the bottom of a building down to street level.
Which industry? 2D animation is alive and well in television animation, for example. You have less control in 3D animation (there are numerous things a 3D character rig has troubles doing that can be done in 2D by, well… drawing exactly what you want). 3D does not take less time… the time is just transposed to a different part of the pipeline. Unless you’re doing something like Roger Rabbit or Cool World, you typically don’t even want to mix 2D with live action. And you really think that there are more 3D modelers than people who can draw? Really?
3D does allow (I daresay necessitates) for greater division of labor in a large project pipe. Where in 2D the division breaks down to a lot of people performing roughly the same tasks (i.e. drawing… keyframe artists, inbetweeners, inking, colorists) and a few people doing technical tasks (projection/photography, final edits, post, etc.), the division of labor in 3D can be broken down in a more granular way to much more specialized tasks that are [theoretically] more modular. If you’re looking at film/television production like manufacturing, then this break down has a lot more appeal. For example, you don’t have to rely on finding/training a lot of artists to draw on model in the same style and convey movement the same way. Your animators can focus on just movement and motion, liberated from the notion of staying on-model… since that’s already defined by, well, the modeler.
3D allows you to more cleanly do animation on an industrial scale. Aesthetic appeal aside, that’s its primary advantage. If you’re not planning on churning out many films with a large team and you want a 2D look, you’re better off just doing the 2D animation with 2D techniques. Look at it this way… if your team is small, you’re going to need more generalists. Since much of the 2D pipe requires the same (or sufficiently similar) skillsets, it’s easier to find 2D animation generalists.
Good animation is good animation 3D or 2D has no bearing on the quality of the animation. you seem to be implying that 3D animation is always on 1s and uses pose to pose with no breakdowns or manual spacing. This isnt the case at all. Also like Fweeb said 2D is thriving in Television.
Disney, Cartoon Network, Adult Swim, BBC are all spending large amounts of money on 2D productions.
I think what your really trying to achieve is a render that looks like it was drawn. This is an up hill battle. the computer doesnt understand style and aesthetics like an artist does and I doubt it ever will. I’ve never seen a 3D render that looked as good as a 2D cell when it was trying to mimic classical styles. ( paperman doesnt count as it was a hand crafted post process)
Big market, movies and such. Control not so much mean in the aspect of the characters rigs, but for the numerous other aspects of 3D.
Being able to draw, and being able to animate are two different things. Though being able to draw does supplement into animation giving the person has interest.
I do see the points you’re trying to make though. Why not just do 2d? Because of the novelty really. I want to see it done and done right.
IF an effective pipeline could be produced for such a technique, then after asset creation, animation would likely be easier than normal 2D as well as require less people.
In a 2D pipeline, You need few animators, and tons of peoples who can draw and interpolate drawings, called inbetweeners.
So yes, being able to draw and being able to animate are two different things, but in 2D, You only need a few animators, while in 3D, everyone who touches a rig has to be an animator, as it is the computer that does the inbetweening!
We’ve seen the industry gradually move away from 2D animation and to 3D, and for good reasons too. 3D allows greater control, takes a bit less time, can look amazing, and can be integrated seamlessly with live action in the forum of special effects. It also has a larger base of people who can actually do it or at least a division in, for example modeling.
I think this is a very misleading …blanket…statement to make. It depends on what you’re producing for example…south park is 2D …that would not take as much time to produce as some of those well drawn classic Disney movies. Also I believe many people watch some of those 3D movies …think it is very kool …and are always in for a rude awakening when they actually realize the effort it takes to put together a 30 second piece animation.
That said …I have Blender and several other 3D programs I bought…I also have the latest Toonboom, a cintiq and looking for a way to capture this type of 2.5D look which I like personally. So this is a good topic for me.
First let me use 2 examples of what I mean by 2.5D:
Look…I realize these are corny…but it serve the purpose for samples.
See, they take those 3D models, do a rather good job of rendering them as 2D, and are sure to animate it as if it was 2D. The lighting was a little too complex though for what I had in mind. ideally, the models would also be a little simpler.
What the hell are you talking about? A hand-drawn animator, even someone who just does the in-betweens, is not just “someone who can draw”. Try finding a job as a hand-drawn animator outside a (north) korean labor camp, it’s practically impossible.
I don’t think it’s controversial to say that the quality hand-drawn animation that you had in the US is dead and gone, and the reason is cost. Sure, if you’re Disney you can do CG films that are as expensive as their older hand-drawn features. Producing cheap CG animation is still much cheaper than producing cheap hand-drawn animation.
Yep the gameplay looks 2d all the way in Guilty Gear. The cut scenes and supers…don’t. They look pretty awful actually and because of the way they were trying to animate them they just look choppy and weird. In that respect I think Naruto Ninja Storm looks much better. However when you are actually playing the game and moving around, it’s freaking amazing how close to the original guilty gear the animation looks.
There was this translated Japanese article going around that details everything they did. It’s surprisingly low tech stuff, like using the classic flip normals technique for the ink outline, all the animation is done pose to pose with interpolation turned off to give it that classic anime look.
The developers have actually always animated their characters in 3D, even with BlazBlue, they would just render it out and send to China for cleanup and line work. The developers behind KOF13 did this as well. It looks 2D but it’s all rendered 3D models with the lines added in post. Anyway with Xrd Sign they decided to cut out all the middle work and go all full 3D, for several reasons. The first, Arc System works are great but they’ve never really used any of the techniques that other developers who make 2d fighter use, like sprite tiling. The actually load the full sprite in memory (and have done so even all the way back from the very first Guilty Gear on there PSX), that’s pretty inefficient and loading those texture maps into memory can get unwieldy especially if dealing with the fidelity that is needed for next gen systems. Even with the huge amount of ram in today’s video cards it’s just unfeasible after a point. Other developers who do use more optimal techniques (like the Skull Girls developers) can push pretty high fidelity sprites on screen no problem, though eventually even they will run into issues when dealing with 4K, if it ever gets to that.
So they went 3D, used unreal because it was cheap at the time (they are using Unreal 3, a lot of the stuff they do like z-ffsetting the characters, may not even be possible in Unreal 4 since it’s a deferred renderer). Ported their fight engine and went to town. Very impressive stuff.
Anyway if anyone is interested the info is out there. I can post the link if people are interested it’s pretty detailed.
Sure. Here is a link to the original article (In Japanese) It’s two parts.
And here is a link to a rough translation someone posted on Polycount. The user’s name is Chev.
This is an interesting article because it’s rare for a Japanese development house to go into so much detail about their process as they tend to be pretty secretive. There are a lot of pictures and videos too which is even more rare.
That whole thread is interesting to read as all the users try to figure out how Aksys did it before the article came out. There are some interesting ideas there if cel shading is your thing. Like using a map to remove detail so that the cel shading is more natural.
Using a normal map to remove detail, who would have thunk. The complete opposite of western game development.
Depending on how the scene is built and the length of the lens, you can animate the “Shift” parameter on the camera to do a fake pan. However if you shift too much in any direction, the perspective starts to distort.
