Compositing
Compositing refers to combining multiple image layers into a single, seemingly real image. Films, commercials, and today also AI-generated videos are almost always created from such overlaid layers.
Compositing means: several images are layered on top of each other until they appear as a single photo. An actor, for example, is filmed in front of a green screen. The green area is later removed, and behind it a painted city appears. To that, one might add rain, smoke, and a flying spaceship. Each of these layers is created separately and only combined at the very end. The result is an image that never existed in front of a camera.
Why almost every modern film consists of layers
Many things cannot be filmed in one go. A city from the year 1890 no longer exists. An explosion right next to a person would be life-threatening. Compositing solves this by separating the parts: the person is filmed safely in a studio, the explosion is created on a computer. Combined, it looks like a single shot.
The second reason is control. If all the components are available separately, each one can be changed individually. Is the sky too gray? You simply swap out the sky layer without reshooting. In practice, elaborate cinema shots consist of dozens, sometimes hundreds, of layers. This is what makes the director’s subsequent change requests affordable in the first place.
It is important to distinguish this from mere image editing. A filter makes an existing image prettier or more colorful. Compositing, on the other hand, creates new image information from multiple sources. The term originates from film technology, but is now used just as much for photos, games, and user interfaces.
From the green screen to the finished layer
The first step is called keying. This determines which part of an image remains visible and which becomes transparent. For a green studio wall, a program handles this automatically, because green is clearly distinguishable from skin and clothing. For complicated edges like hair or smoke, this isn’t enough. Then people trace the outlines by hand, frame by frame.
Technically, transparency relies on an additional channel, the alpha channel. For each pixel, this stores not only color but also a value for opacity. Zero means fully transparent, the maximum value means fully opaque. Intermediate values create soft edges. It is precisely these intermediate values that determine whether a composite looks real or like a cardboard cutout.
The hardest part is matching. Two layers only appear to belong together if light, color cast, sharpness, and image noise match each other. If sunlight falls from the left in one layer and from the right in the other, the eye immediately notices the deception. That’s why compositing software often adds artificial noise instead of removing it. AI tools now help here too, for example with automatic keying without a green screen.
Compositing in streaming, video calls, and AI videos
Most commonly, one encounters compositing in video conferences. The swapped-out background is nothing more than a composite in real time. The software recognizes the person, makes everything else transparent, and places a different image underneath. The fact that hair sometimes flickers in the process shows exactly the problem of clean keying.
Weather forecasts on television and the offside lines displayed during football matches are also created this way. In news about film productions, the term usually appears together with costs: compositing is one of the most expensive line items in major productions, because a lot of manual work remains necessary.
In the AI industry, the topic is current for a different reason. Video generators often deliver a finished image that can no longer be broken down. But studios still want to be able to edit individual layers. That’s why providers are working on generating AI outputs directly as layers. Anyone wanting to understand why generative tools haven’t replaced everything in the studio yet will find the core of the answer here.