The process of censoring objects within a video is a post-processing operation that takes place after the video has been recorded. Censoring can be effectuated using any effect that makes the object to be censored barely visible - this ranges from using a blurring effect and up to just a solid color, both of which can be overlaid on top of the objective to be censored. In case the objective moves, the difficultly in censoring consists in manually applying adjustments using keyframes for the duration of the video clip to ensure that the objective remains censored as it moves.

The recipe to censor objects within a video consists in:

• Duplicating a video track and placing it as an overlay above the original video track,
• Censoring the video track copy using an effect that renders the video track copy difficult to understand, for example, by using a blurring effect to make the scene barely visible.
• Use a crop effect together with key-frames in order to crop the censoring overlay down to the size of the objective to be censored.
• Using keyframes and motion-tweening in case the area to be censored is moving.

The habit of the trade is to use a blurring effect that makes the objective barely visible, however, as mentioned a black block is also perfectly valid for censoring, with the difference between the two being that the blurring effect provides more contextual information about the objective and the area around the objective, which might be necessary depending on context.

For example, imagine having to censor all individuals in a video clip during a shootout where it is preferable to see the guns being fired but it is additionally necessary to censor the people firing the guns. In the former case, a big block of a solid color would not provide the information that someone is shooting, such that just blurring the individuals a whole is a better strategy.

The task is to censor the "stranger" from "Color Classics - The Little Stranger" cartoon (produced by Fleischer Studios for Paramount Pictures) for a sub-segment of the entire original videoclip (for brevity).

with the following project files:

• censoring.zip

Here are some observation regarding the realization:

• It is possible to use crop keyframes just partially for "top", "bottom", "left" and "right" but it might turn out confusing for the whole process such that it is easier to just punch in all the crop directions.
• The realization is done using a "rolling-shutter" approach where the entire videoclip is adjusted as a whole and progressively instead of creating slices and sub-clips which might end up more confusing than helpful in the long run (the impulse is to slice up the clip).
• Highly dynamic sequences, mostly in terms of velocity, are adjusted frame-by-frame. For example, when the duck picks up the "stranger" and sets it down. This sequence is manually adjusted frame by frame, as opposed to allowing the computer to "motion tween" (transform) between two keyframes, due to the movement possibly exposing the objective.
• In general, the formula to get the censoring done, is to just go through the entire clip, perhaps even using multiple passes and whenever the objective is partially exposed, adjust the censorship overlay video clip to cover up the objective. In that sense, it is not difficult to do but just requires work.
  • Ideally, the clip would be offered to someone else that is neutral, to proof the censoring.
• For upcoming technologies, artificial intelligence (A.I.) might be a good avenue to explore for the software creators (Adobe, in this case), to perhaps provide a way to isolate the objective from the rest of the environment. Given the limitations of what is called A.I., the results will not be perfect, but might cut down on the work that needs to be dispensed making the blurring rectangle narrower throughout the video.
• As an observation, ironically and funnily, due to the clip consisting in an older cartoon, the dynamics of movement are sometimes relative to the absolute frame, which means that the movement ends up holding the objective in place, such that it is easier to censor. This can be seen in the last sequence when the ducklings trot off on the left-hand-side of the scene, where no adjustment to the censorship frame is needed because every object holds its place (probably individually animated and then placed onto the scene). This makes adjusting the censorship frame much easier.
• It is a bad idea to try and make things tidy and symmetric, which ends up having to use more keyframes and adjustments as well as potentially missing some frames where the objective might be exposed. Ideally, if the job is done, do not make it additionally cute.
  • For example, at the start where the "stranger" fades in both left and right depending on the upcoming scene, the censorship frame is extended as a rectangle. This is alright. You have to keep in mind the objective, which is censorship, not making things look cute or even.
• The size of the censorship frame hinges on the amount of context necessary to be exposed and results in less or more work depending thereon.
• Dynamics for cartoons are more than often drawn using mathematical patterns. For example, the sequence where the "stranger" is picked up and moved is almost a perfect bell-shaped or semi-circle movement. This requires mostly a few keyframes, such as: at the start, at the end, at the peak of the movement and then depending on the content two additional ones to keep the censorship on a curve. Very similar to integration. The same applies to repetitions where character animations are just repeated multiple times (ie: the large duck shaking multiple times to show disgruntlement) such that adjusting the censorship frame is not necessary (or can be applied the very same way as the preceding sequence).