Comment gérez-vous les occlusions et les intersections dans le roto?

Always build layers starting from background to foreground. Comps are always built in this way as should roto. I have qced multiple rotos, usually hands/arms where consideration is not taken for objects created with one spline/layer that are both over and under another spline/layer. A good example if a hand is clutching another object. There needs to be a layer for all the bg hand splines. A mg layer for the object being clutched. Finally, a third layer for all the fg hand objects going over the mg and bg parts. Sometimes fingers need to be created into individual layers as they may need to switch from going over and under a clutched object frequently. A good practice is to study the shot before creating the intial roto.

I feel like these ‘ expert’ answers are just here to train AI 🦾 101001010010100100101001001010101010101010101010101010101001011

My best practice for layering roto is 1) ask what purpose does the roto serve in the composite then 2) layer accordingly. For example, let's imagine the Comp needs an actor's performance from one plate placed on another background plate. Then, the roto artist must roto the silhouette of the actor and only one layer is needed. However if the goal of the comp is to make that same actor's eyes glow, then the comp requires two roto layers miminum: the eyes and anything that occludes the eyes. Recap: Roto should be layered according to its purpose(s) in the comp.

When rotoscoping humans, it's a common practice to simplify the body into natural anatomical shapes for greater accuracy and ease. Essentially, this involves outlining and identifying the different parts of the body like the head, arms, legs, feet, torso, hands, and even down to each finger, and then animating them individually. By doing this, it guarantees that the movement of each body part looks natural and lifelike, while also making the rotoscoping process more efficient and precise. By following this method, you can position your pivot point at the natural joint locations, and then rotate and move the shapes accordingly, rather than making significant modifications to the shapes themselves.

The practice of pulling out the feathered edge of a roto spline is generally avoided in my studio. And I hear other studios also avoid it. The reasons are: 1- that use of the feathered edge doubles the number of roto edges to manage. 2- the feathered edge slows roto software performance. 3- the feathered edge hides wobbling, incorrectly placed roto and is more difficult to edit The Prefererred Method for Feathering the edges is to use a separate procedural method. Nuke has a one button motion blur feature which we can trigger at render time. Turning on this feature only at render time preserves the roto software's speed.

I have also noticed a lot of roto being produced where shapes are being turned off/on when they go behind another layer or as soon as they reach a point where they are not over a layer behind. There is no good reason to turn off/on shapes in these instances. More often, the shapes are being turned off too early and get kicked back to have the shapes turned on and rerendered or done by the next artist upstream delaying the process. All shapes should remain on until the entire object being rotoed has passed the area needing coverage. Selectively, turning off individual splines once they move beyond the area required reduces the layer ability to be used as entire object that may be needed for other functionality, usually color corrections.

I agree that holdouts and any occlusions to the main roto layers should be separate rotolayers. I also agree that these holdouts should not be subtracted by the Roto Artist. They should be roto'd as positive mattes. The Roto Artist should roto all these targets in separate layers. Any subtractions, or other channel operations, should be done by the Compositor.

A holdout mattes work very well if you have an occluding element. If it's a simple holdout and your software allows multiple shapes, including negative shapes, you could do it as defining layers. One advantage of keeping it as a separate holdout is it independent of any changes to the actual key roto such as adding more characters or revisions.

There are two schools of thought about roto and keys. One group believes Keys should not be mixed in with Roto. The other group feels its ok to mix any technique to achieve a complete matte. The first group - "Roto only" Group believes that many compositors prefer to do their own keys. So many Roto Artists don't key because the Compositor will throw out our keys anyway. This "Roto only" mindset is usually found in big studios. The second group is the "Key & Splines are ok" group. This group is usually found in smaller studios where the Compositor also does their own keying and roto. I can operate with both strategies but I prefer the "Roto Only" group. I philosophically believe "Compositors expect Roto as splines. Deliver accordingly."