A Revolution in Compositing
Deep compositing, or ‘Deep Image’ compositing, is a groundbreaking technique that fundamentally changes how computer-generated elements are layered together and with live-action footage. In traditional 2D compositing, each pixel has only color and alpha (transparency) information. In a deep image, each pixel stores a variable amount of information; it contains a list of color, transparency, and, crucially, depth samples for everything the camera’s view intersects at that pixel’s location.
Solving Traditional Problems
This extra depth information per pixel solves many long-standing problems that were incredibly difficult and time-consuming with traditional compositing methods:
| Problem | Traditional Solution | Deep Compositing Solution |
|---|---|---|
| Semi-Transparency | Complex, multi-layered mattes to place smoke or fire behind a character’s arm but in front of their body. | The deep data automatically knows where the smoke should be in 3D space, correctly rendering it around objects. |
| Holdout Mattes | Manually creating ‘holdouts’ or garbage mattes to make sure objects appear in the correct order. | The depth information inherently sorts all elements correctly, eliminating the need for holdout mattes. |
| Re-rendering | If an element is added or changed, many other layers might need to be re-rendered. | New elements can be dropped into the deep composite and they will automatically settle into their correct depth position. |
Developed by Weta Digital for Avatar (2009), deep compositing has become a standard for complex, high-end visual effects work. It results in a more accurate, flexible, and efficient workflow, especially for scenes with intricate geometry and volumetric effects like smoke, fog, and water.