Графика в Futurama'е рулит. Цитата by Scott Vanzo:
To clarify for anyone curious, we use Softimage Toonz for 2D ink&paint/composite, and Alias|Wavefront PowerAnimator and Maya for 3D...
Полный текст:
I agree with mArc's observations regarding the lack of artifacts in 3D animation that cause it to diverge with the quality of the hand-drawn (2D) animation.
Hand-drawn Animation for episodic (television) production is generally animated on 2's(12 fps) for any particular movement. This cost-effective measure introduces temporal artifacts such as strobing and emphasizes video field separation due to the NTSC format and 3:2 field rate conversion(converting 24fps->30fps).
These artifacts are generally considered limitations, so the prevailing attitude among our episode Directors is to use all 24 fps of 3D animation, despite the disparity. IMHO, however - the 12 fps would be my personal preference UNLESS the motion was considerably fast or complex. Even hand-drawn animation is animated on 1's when clarity is needed or a camera pan is in effect. Perhaps a more judicious use of 1's would be more appropriate.
Spatially, a hand-drawn character is often "cheated" to emphasize traits of the face or body - to improve legibility of an expression, recognition of character(through silhouette features), and/or design(composition). The character's features are often drawn slightly out of proportion or off-perspective as a result.
Bender is a good example. His visor is clearly visible in hand-drawn rear 3/4 poses even though his visor is all but occluded by his head in our 3D model. We would have to severely lengthen or warp his visor during a 3D animation in order for it to appear identical to the 2D model. Also, front 3/4 drawings emphasize the frontal features by "pulling" the front of the character more towards the viewer. All of this pushing, pulling and warping that is merely an intuitive impulse/response to the 2D animator would be a logistical nightmare to the 3D animator.
This problem would be interesting to investigate, however. A more procedural way of key-framing these distortions, based on camera movement, may be possible.
Unintended variances in hand-drawn perspective and position are often a trait of even good Animation. The eye is able to "accept" some imprecision without really correcting for it. The movement is believable, even if it isn't incredibly smooth. In the worst case, an object that is drawn very small, or in-betweened poorly may jitter unacceptably.
This is where we could experiment more and still meet our schedule. Perhaps a minute, random offset, rotation, camera angle adjustment could be introduced to each frame to vary the appearance in a subtle way. This would not be too complicated. Worth a test.
3D is used to extend our 2D toolset, so its use is by definition divergent from what we can do with 2D. We therefore limit our 3D animation by making the movement as simple as possible.
?Drawn high-lights and shadows on a character or object may be based more on composition than on accurate representation of a "3D" or real-world light source.
In general we "lock" down, or limit the way that lights interact with objects. Usually, it is as simple as grouping a light source with an object, where possible, to reduce the broad movement of shadows or highlights. The light source can move locally, but not globally.
The amount of coverage and softness between shade regions is also considered.
?Line thickness and coherence(missing lines or line portions) is an issue for hand-drawn Animation (during the Assistant/Cleanup stage) as well as for the 3D Animation. We've had to learn how to model in 3D to help insure that lines won't jitter or pop on/off during Animation rendering. Our rendering solution(PowerAnimator->PowerToon) was the best product at the time to maintain coherence in lines and still offer a great amount of control. We're still looking for a Maya solution for line rendering that will offer us this kind of quality.
Line quality is another issue. The line renderer we use is good, but creates a uniformly thick line. This is adequate for Futurama, but there are some finer nuances of 2D cleanup that are difficult to emulate, such as rounded corners and soft pencil lines.
Effects (such as smoke, explosions, nebulae) are toned down quite a bit so they don't look too realistic. Many people think that cool looking effects are necessarily realistic-looking; I contend that abstraction can look appropriate, if not "cool", as well. Abstracting effects (reducing local detail/noise, lighting and action) is much more difficult to do well than pursuing a quasi-realistic result. Of course, obtaining convincing realism is a holy grail of sorts, butWe have to determine how far we will go and still integrate the style with the hand-drawn. Some shots are more successful than others; Overall, IMHO, we've done pretty good.
-----------
To clarify for anyone curious, we use Softimage Toonz for 2D ink&paint/composite, and Alias|Wavefront PowerAnimator and Maya for 3D...