VFX_Rotoing your heart out!

One of the essential skills that a VFX artist should have in their tool kit is rotoscoping. The position of Roto Artist used to be the first stage in the career of VFX compers, although nowadays this kind of position seems to be disappearing from the London job market. Nevertheless rotoscoping is an essential skill that all artists (doesn’t matter their level) will one way or another use in their everyday jobs.

Rotoscoping was intensively and effectively used in a number of later Fleischer cartoons, most notably the Cab Calloway dance routines in three Betty Boop cartoons from the early 1930s,

Rotoscoping is an animation technic where the artist will create several shapes that will mimic the movement and shape of an object creating, as a result, a mask or matte of the element. This new alpha/matte will facilitate the manipulation of that object in the shot with techniques such as extraction, grading, overlapping, etc. This technique is not unique to VFX artists, it’s also a method used in other stages during postproduction.

A few of the following observations are useful tools for becoming an efficient roto artist. Tools that not only will help you to roto but that you can extrapolate for other comping technics.

Roto is a time-consuming process, so before you start creating any shape, you should come up with a plan.

First, you need to check your shot. Are you working in the right range of frames? Have you isolated the correct elements on the plate? Do you know the objective for the roto? The fina, use it will have? Once you have checked what needs to be done and if the plate is correct, you should watch the footage a few times and start analysing the element that you are going to isolate. How the object moves; do you have motion blur; do other objects in the shot occlude it? what about the edges of your element, are they blurry, wrinkled, sharp, etc.? At this stage you will be training your eyes and brain to visualize how to attack the task of isolating the object.

Now it’s time to create some shapes! The usual rule of thumb in the roto world is to divide and conquer, keeping to one shape at a time. The idea is to break down our object into multiple simpler shapes, one shape at a time. We create one shape for a particular section in our object, we follow it, and then we go back to the beginning and create a new shape for another section. This process will allow us to have more control and keep consistency over the whole shot. By creating smaller shapes and keeping the look consistently despite the movement, scaling or rotation, we will avoid jittering, boiling, deformation or odd movement around our edges.

Rotoscoping isn’t only about the shapes of our objects; it is about the object moving through time. Therefore we have to talk about keyframes. I believe the best result is achieved by a motion based roto where we only create keyframes when our objects change direction, or the movement is stopped. The previous analysis of the object’s movement will help in the creation of strategy keyframes. The software will generate movement and positions between the keyframes; this is called interpolation. This is the process where the software will create positions between user-defined keyframes. This information will be an average of the bordering keyframes.

To take full advantage of this process it’s necessary to use as few keyframes as possible. To achieve that it’s sometimes required to generate a tracking.

Tracking is a time-saving option. The basic idea is to generate as many tracking points from the object as possible to use this information to reduce the amount of movement of the object, limiting the number of keyframes needed. In this way we can follow the motion path of the object creating keyframes based on this pattern. You can link this information to your roto node, or you can stabilize the motion of the plate to generate the roto shapes and then incorporate the movement afterwards. (don’t stabilize the whole plate to match move later as that will filtrate the image and ruin the plate).

There are a few things that are good practice for compositing in general but good to learn when you start with rotoscoping. LABEL YOUR THINGS!!! You will be sharing scripts and comps every day, no one owns a comp or branch of the script. So be kind to yourself and your fellow workers and label your nodes and branches on your scripts. Backdrops are a good idea, also some people like to label every shape, which sounds a tedious task, but if you are roto-ing a complex shape, it will pay off in the end. Also if you have several shapes that belong to specific sections, e.g. you are rotoing a human head, and you have various shapes; organize them in various identifiable groups: a section for the ears, the hair, the mouth, etc.

Once that’s finished it’s essential to double-check your work carefully. Remember the objective is to create a cohesive matte, consistent edges, no holes inside your object, no overlapping of shapes (they might create unflattering junctions), etc.

For best results, I would recommend to roto by overlaying your matte in colour. E.g. you can use a background in a mid-grey colour. This will make any inconsistency more evident that reviewing it over a white or black background. (Some times those colours can give you misleading results). Also, it will become evident if you have edge inconsistencies between shapes or if your shapes don’t move as one cohesive unit. It’s okay to have inaccurate shapes as long as they are all consistently inaccurate in the same style. If your shape is 2 pixels away from the edges, all your other shapes will have to follow that rule. In this way, if you have consistently inaccurate shape/edges they can be manipulated at the same time by a erode/dilate effect.

Final tip: toggling between various ways to view your matte is an excellent way of viewing the accuracy of your mattes.

---

Bibliography

https://cinema.usc.edu/spectator/36.1/4_Yang.pdf