next »

[Hangedman]

I think learning the tools for 3D art is considerably harder than learning the tools for 2D (and 2D tools are more easily accessible and less complex to use in general, workflow etc).
I think animating well in 2D art is more difficult just for sheer amount of work and adjustments that need doing (no rigging, automation, etc, except with bone-joint 2D animation which takes forever just to make not look like crappy puppets).
Once you know how to do either well, I think it's pretty much even. But I don't really think it's really worth comparing in terms of time commitment after it's generally mastered because the time is used for entirely different things in one as opposed to the other.

As an aside, I'd like to take this discussion down an application side path, if anyone cares to come with me.
If not, feel free to disregard this.
I'm thinking about my next project after my current ones. A bit more ambitious. I want to make a stylish beatemup, a 'spectacle fighter' in the vein of Devil May Cry or Dust: An Elysian Tail.
Originally, I wanted to do it in 3D. This would be closer to my original vision, more practical for big spectacular fights, and would allow me to get away with simpler art (or rather would let me make it as detailed or undetailed as I wanted). But to do so would require me to learn Unity and a decent 3D model making app, plus all the nuances of 3D.
2D art would be much simpler initially, but animating it could take forever, and I wouldn't be able to let myself off with simple art. And it isn't quite what I want out of the project. I can't even be sure which would be easier. As I said, both require an equally serious time commitment, but for entirely different reasons.
In essence, this whole debate is going on in my head at the moment, and this thread has yet to help one side win. Any thoughts?

[Geti]

Hangedman considering you're going to need a bucketload and a half of animations for people whacking each other and you're probably going to want to tweak them all over the place 3D with skeletal animation and all would likely suit a beat-em-up more. That and you can zoom in for super-KOs and swing the camera around wherever you like.

If you've never done anything in 3D before it'd be a good idea to get some practice prototypes done first though

[gimymblert]

Make 2D concept art/collage to show graphical target and then tweak the style/target according to constraint/tone.

Also the intricacies of 3D is mostly straightforward, you need to research the correct techniques though (Hint: modeling JOINT is the first thing to document for 3D animation, especially crotch, shoulder and limb. Other thing like edge loop and edge flow are related and important to understand). You can live entirey on subdivision, extrusion, welding vertice. UVMAP is made easy on blender so it is mostly a non issue, and you can directly paint the model and uv to rough something very quickly. You can use also trick with multiple UV projection cloning to bake afterwise in the main "game UV" with blender to ease Uv painting into trivialness.

Rough the game with low poly model too before upgrading (no need necessarily for skin mesh, but old school aggregation). Don't forget texture are pretty powerful in the final visual quality, shader had even more layer of wow, sometimes more so than number of polygon if you have good and expressive silhouette.

edit:
I don't now about the new version of mainstream 3D software, but bone are never baked at skinning phase in blender. That mean you can still edit mesh after skinning (you can't in other soft), so you can have a base topology and base skeleton (armature) edited separately (although you need to reassign any new vertice created through extrusion, also allow you to quickly edit rough character build and keep the vertice group in the process, allowing you to skin a base model only once, work great if you have a 3D paperdoll creation of character, no need to reskin new model). Right now I only train with low poly 1 vertice per bone to have good practice of low deformation placement of bone, but I abuse this to experiment.

[Emme]

This is a very interesting discussion here, and it happens a lot in our team, too.
My opinion on the subject is that it depends on the visual direction and the actual amount of content AND the genre.
After years of experimenting I can say that its possible to work around the effort of say hand pixeling characters (and animating)
by setting up a model and apply the right shaing and post-work.Then You can make it look very handmade and still use all the tools
in the arsenal to make the animation look good.Once You made one model You can quickly add variations by simply changing colours, shaders, and quickly modelled assets.
The downside is that 2d art scales not so well.On the other hand, games like Cthluhu saves the world or Avadon dont seem to suffer in the public's eye just because they dont scale well. 2D art is BVERY quick when You need to make changes.Prerendered art can quickly and easily be manipulated in a paint package to change the look and level of detail.

Now as for 3D:it scales well and reacts to lighting, and thats already the best I can say about it.
The workflow for state of the art games is tedious:
1) Build basemesh
2)Export mesh for scultping in Mudbox or zBrush
3)Export sculpted/textured highpoly mesh
4)Create lowpoly mesh
5)Lay out UVs
6)Animate mesh
7)Bake normal, cavity, ambient occlusion, diffuse, specularity (of course some of them have to be created first)
8)Tweak

Doing corrections, changes etc in 3D models can mean that You need to repeat the process from step 2).
Personally, to me 3D makes only sense if you use a reduced visual style, OR have a team of artists (more than 2) who also
are experts.

[gimymblert]

But state of the art is always difficult, it make no sense to pit average case vs state of the art (ie not pixel).

And in your case it is still about still image. While you might tweak the mesh, the animation will most likely remain unless you make major change, and even there you might save your ass by adapting the actual animation instead of redoing everything.

If you have no 2D expert it's equally difficult to have good visual quality and change are even more expensive (need to toss everything unlike 3D you don't start at 2 you go back at 1).

[Emme]

"And in your case it is still about still image. While you might tweak the mesh, the animation will most likely remain unless you make major change, and even there you might save your ass by adapting the actual animation instead of redoing everything."

Hm, could You explain this again, please ?

[gimymblert]

Speaking of character animation: male shepard/female shepard in mass effect share the same animation, sometimes it comes out odd because it's only male but it does save time, also heavy customisation.

That qualify as mesh sharing same animation base. Notice how the facial animation also have some same quirk with lipsync (automated) and the famous eyebrows raising all character share across even alien mesh. Alien or character does have their own set of animation, but some are generic retarget and batch across all model. The technique have been extended for swotor greatly (Entire scene are batch through automation)

This is nearly impossible in 2D with that level of flexibility.

Here is some talk about it:
http://www.gdcvault.com/play/1014362/Cinematic-Character-Lighting-in-STAR
http://www.gdcvault.com/play/1012460
http://www.gdcvault.com/play/973/Universal-Character-System-in-SAINTS
http://www.gdcvault.com/play/1709/Style-in-Rendering-The-History
http://www.ecgconf.com/videos/view/tgc-2011/3960-Facial-Rigging-Part-1-Pending-Review
http://www.ecgconf.com/videos/view/tgc-2011/3961-Facial-Rigging-Part-2-Pending-Review

Notice it is for AAA quality, you might not need something as deep with wind waker like animation and lighting.

On animation sharing if you are not aware: First I use blender where bone and mesh are independent It's easier to share, swap, modify data. First on every program you can decouple animation from a base skeleton and reapply animation (on the modeling soft or in game engine) from a base bone setup to any model who share the same hierarchy. Also you can break animation in part and blend them together, for example so you can combine reloading with walking/running, etc...  I don't know about other program, but in unity it just a matter of reapplying an animation clip (extract at export time) to another rigged mesh. Using the same animation across different proportion or different bone setup is called retargeting, most 3D application have some of this, I did not research this in depth. Similar hierarchy with the same bone naming would still work, but they might be odd if the proportion are off for some animation (walking might be share able if you modify the size or shoulder size, but hugging might suffer from it).


If SOME indie consider such technique early on in their workflow, they will end up with a great tools which will quickly offset the initial cost. It's well worth the effort in the long run.

If it's a problem with artist who want to remain in control and avoid generic animation, you can still class character with a tier system and keep all generic animation for low tier generic peon, and keep all the artist exception for important top tier, and use a mix of the two with middle tier (only have a small set of specific animation which might be blended a bit with generic animation).

A good universal and flexible model does take some expertise and time to build, but the basis is rather straightforward and have big ROI.

[oahda]

hiring a competent 3D modeler/animator is on the average much more expensive than hiring a competent 2d animator/pixel-artist. this is a fact.

you can build a 2d game off free tools and suffer no consequences. building a 3d game from scratch is much more difficult, and using third party engines and tools like unity and 3dmax can be very expensive (especially if you wanna go commercial). getting a 3d game to look good in this day and age is a lot harder than getting a 2d game to look good.

don't turn this into some gayass discussion about 2d vs. 3d game development. Just know that I'm right and you're wrong, argument ends here. Want to make a new thread for it? Go for it, don't derail this thread further.

(fucking splits)

Having experience with both Blender and Maya, I will seriously tell you that Maya sucks for modelling. Really bad. And I've used Maya 2009, 2010 and 2011, both on Windows, Linux and Mac. It's crap for modelling. Nice for animations and simulations. Shite for modelling.

Blender, on the other hand, I will honestly tell you, is the best I've ever used for modelling. So many hotkeys (Maya has like five useful hotkeys; Blender has them for almost anything), much better tools for symmetry (Maya barely has them), getting rid of unwanted geometry, filling out empty faces and so on. There is no reason not to use Blender for professional modelling.

[gimymblert]

There is no reason not to use Blender for professional modelling.

Tradition, school and lobby

[oahda]

There is no reason not to use Blender for professional modelling.

Tradition, school and lobby

There is no good reason, then.

[Emme]

Maya is excellent for modelling. As is Max, XSI, LightWave and Blender. Its just a matter of exercise and personal UI config.
The best thing with Blender is its free, feature complete and powerful, which is right for indie game development.

[Dr.Electro]

maybe i missed it in the discussion, but:

3D has the advantage, that it is resolution independent.

of course the textures should be otpimized to fit a certain resolution, but it doesnt really matter if you go for 480x320 or 550x320 or 1200x800 ... try this with hand pixeled graphics. (yeah i know vector graphic is fine too )

i prefer working with 3D, but i would say it should depend on the game you want to make, rather than say "this is it".

[J. R. Hill]

I bet in years past renaissance artists had discussions about whether painting or sculpting was more difficult.

[gimymblert]

http://www.sculpture.net/community/showthread.php?t=2864

Michaelangelo and Leonardo had such an arguement 500 years ago
The focus was as to which was the more noble art

[J. R. Hill]

I was thinking ninja turtles when I wrote that too lol.

[xrabohrok]

I can speak to experience on this topic.  During a school project where we were working on a top-down shooter, our team ran into an art bottleneck.  Specifically, animation.  Our artists, both of them and sometimes me, were trying to draw these things in paint.net.  This takes a LONG TIME. The programmers then decided that they would be running the animation at 12FPS, which translated into a lot of detailed, hand-drawn frames.  Once they were in motion, they wouldn't look right.   The scope of the project was out of control.

  I had some experience with 3D before, in fact I owned a weirdly licensed version of 3DSMAX 7, but I had to use something else.  Thats when blender came into play.  It was my first time learning the controls, but once I spent that week or so making a monster, it was considered done.  If the FPS changed, I rendered at a different FPS.  If the programmers added an action to the monster's repertoire, I just added the action and rendered that.  If the sprite formatting changed,...you get it. 

  The artists where then left to draw the settings and more static things, things that are finished once just one are drawn. 

The moral of the story is that you aren't looking at a choice of whether or not one is easier or not: I can model a passable character in a hour, and have it ready for whatever in another. There are trade-offs: any artists can draw but there is an additional learning curve (kinda steep, admittedly) to get ones that model.  Drawings, once they are done, are done sooner, but are fragile in the face of change. Drawings will need to be redrawn if something doesn't work out. 3D takes longer, much longer sometimes, to get to a usable state but can be fiddled with and tweaked, rendered and re-lit til it is near perfect. 

I've found Blender to be faster and 3DS MAX to be better for complex things.  I also found MAYA to be hella obtuse, interface-wise.