What sort of limitations are there to videogame animations? I've noticed even in a lot of more recent games that character animations feel oddly lossy. Compared to feature film animations, videogame characters have relatively mechanical movements. I understand that there are some limitations for how many animations you can have and how seamlessly you can transition between animation, but there's just something missing. What is it that makes game animations so inferior to film animations? Is it simply the amount of talent and effort that goes into the game industry versus feature films, or are there technical limitations to the amount of motion data that can be stored and played in a game dynamically? How much information can be stored and played? Why are facial animations so often a transition between states rather than smoother movements?

Quote: What is it that makes game animations so inferior to film animations?Is it simply the amount of talent and effort that goes into the game industry versus feature films, or are there technical limitations to the amount of motion data that can be stored and played in a game dynamically?

A game is a real-time simulation. A film is a pre-rendered simulation. Because games are a real-time simulation, games simulate and render each frame in under 16 milliseconds (60 fps). Most games have to sacrifice quality and precision for the sake of speed, and in most cases, the difference in quality is negligible. Films do not have such a tight time-constraint, and are free to simulate as precisely and accurately for as long as its deadline allows.

Quote:
How much information can be stored and played?

As much as the game has budget for.

Quote:
Why are facial animations so often a transition between states rather than smoother movements?

Either that the game cannot budget details in a character's face or linear transitions are computationally less expensive than a smoother approach.

animation in a film might have just been built as a stage would instead of an open world with focus only on what is needed to be seen from the light and camera angles used in the script. Like he said there's no need for speed at 60fps in an open world with angles other than the one you have in the script, as there are no player controlled cameras in a movie, with constanty changing player files.

Just 1 of those frames in a film may take a long time to render (to slow to play a game with) and probrably went through an editor and Had FX added after that..when all is said and done then ya are watching the movie. It may be on a much bigger screen than people are used to playing games on, but no one is going to turn a third person camera while watching a movie and fire weapons off the set..

I know there are constraints but I want to know what they are. I do a little 3D work but I don't know much about how it's stored and what makes motions smooth. I'm not very technically informed but I would like to know what sort of differences there are in, say, the 3DSMax format and whatever format they use in the Unreal Engine.

Quote: Original post by Portugal StewI would like to know what sort of differences there are in, say, the 3DSMax format and whatever format they use in the Unreal Engine.

Probably none. As has already been explained, it's how you use the data, not how the data itself differs.

Quote: I know there are constraints but I want to know what they are.

You've already been given the number one constraint: having to render an entire frame in less than 17 milliseconds. Another would be that you're often not using canned animations but rather blending/tweening them on the fly as the animation of a player-controlled character is dependant upon user input, whereas the movie director has already defined exactly what moves when and how. Therefore it's often not been animated entirely by hand, but rather the software is mixing multiple hand-animated sequences together in realtime. Naturally, computer mixing is not going to be as perfect as hand-blended sequences. The animator can take a a few hours to put together the perfect blend, whereas your game has a fraction of a millisecond to do it.

[Edited by - sybixsus on September 29, 2009 7:55:16 PM]

The constraint is time. 17ms is not a long time, and there is a lot to do in it. Consider than animation will be budgeted at only a fraction of that. Say, 3ms.

Animation skinning has to:
1) take every bone, and blend it towards the next frame of the current animation.
2) take every bone, blend it towards the current frame of the next animation.
3) take every bone, blend in secondary animations, programmer bone controls, and run IK or other tasks.
4) take every vertex and transform it by a weighted blend of all bones that the vertex is weighted to.
5) send all the transformed vertices off to be rasterized

That is a LOT of matrix and vector math. Compounded by the fact that it is mostly random access, as it is hard to organize bones/vertices in a decent access order that is effecient for all steps of the process.

The process can be done on CPU or GPU, but either way it requires a lot of memory and memory access, because you need a copy of the vertices for each rendered animated character.

Offline renders are less limited by all of this. You can render a single character at a time, and composite them all together later. You aren't limited by memory constraints (ie you can build temp files with parts of the scene to render). You aren't limited by processor power of one computer (render farms are setups of dozens of servers all working to render the movie).

All this means that game characters tend to have less vertices, less bones, and less bone weights per vertex in order to speed up processing and reduce memory.

Besides the technical limitations outlined above. Animation for film is largely character driven. Typicaly each main film character will be given to a different lead animator. This animator then, much like a actor, develops the character through motion. The best film animators "act" through animation. And everything is done on a shot by shot basis. That is the animation is continuly refined over and over again until the film director is pleased.

You can get a sense of this through the special features on the Star Wars Attack of the Clones special edition DVD. As they continuly revise one particular shot of Yoda until they get something George Lucas is pleased with.

Comparing in-game animations to film is very apples vs oranges.

You can break animations in a game up into two categories - dynamic animation (in-game stuff, running, jumping, shooting) that is blended at runtime, and static animation for cut-scenes which is simply played like a film.

Film animation is the same as game cut-scene animation.
The animators have complete control over the characters, and motion-capture can be used almost completely. This animation data can be huge, but that doesn't pose a problem as long as you can stream the data in as it's needed.

Dynamic (gameplay) animation is different though -- the animators have to animate lots of small actions, which the programmers then stitch together to create the in-game motions.

With the static approach, the animator can tweak fine details such as where a character will place their foot as they navigate a doorway.
With the dynamic approach, the animator's have to make a huge range of mini-animations hope that the game chooses the right ones at the right time.

For a good example of the contrast, look at any of the GTA games (from GTA3 onwards). Their dynamic animation is normal video-gamey stuff (GTA4 does lots of blending/rag-dolling), but their static animations are specifically crafted for each cut-scene, just like a film.

I already know the limitations of tweening dynamic animations, I was mainly interested in how limited character animation is in realtime rendering.

I would like to know, MSW spoke of the difference between the amount of work that goes into film versus game animation, is this because of the technical restraints of games, or can I owe as much of it to simply inferior animators working on games? I don't want to diss game animators, but I can imagine that the videogame industry doesn't attract superb animators as does the film industry.

I especially am convinced of this because of how much character I see in many Japanese games compared to most American games. I saw the trailer for The Last Guardian recently and was astounded with the personality that went into the character animation. I've seen much less in the way of 3D animated films from Japan but the game industry is packed with excellent animation.

You don't seem to be catching on to the very basic, fundamental difference between animation for an animation, and animation for a game which has a million totally unrelated requirements that need to be taken care of.

Of course game animation is going to be lower quality; ultimately it's one of the least essential parts of a game. You can make a game without any animation in it at all and you can make a lot of games with very basic animation. Now try making an animated film without animation... doesn't work so well, right? Animation is the entire point of an animation. Animation is not the point of a game, so it takes a backseat to the gameplay design and programming which are far more pressing concerns.

Now if you're a studio that has a very large budget and is trying to make a visually impressive game with great animation as a selling point, like say Naughty Dog who just put out Uncharted 2, you're going to focus more on animation and it is going to compare to film. Otherwise, it's just not a logical priority.