Separate image layer encoding distance from camera for each pixel — allows depth-based effects, focus pulls, and atmospheric adjustments in post without reshooting.
You need a separate image plane that stores for each pixel how far it is from the camera — that is your depth channel. While your RGB image carries the color and brightness information, this channel provides you with pure distance data in grayscale form: white = near, black = far. On set or in 3D synthesis, you generate this information in parallel with the final rendering, and in post-production, you have an invisible control map that gives you control over spatial effects.
In practice, you use the depth channel for several things simultaneously. Firstly: depth of field / defocus — you can keep a specific distance zone sharp in post and selectively blur everything in front of or behind it, without having to re-render the original. Secondly: atmospheric perspective — the greater the distance, the more haze, fog, or color shift you apply over the pixels, which suggests depth. Thirdly: keying and compositing — you can isolate elements by their depth to layer complex compositions without manually masking individual layers. When working with multiple render pass layers, the depth channel becomes your navigation tool between layers.
The biggest challenge: The channel requires high bit depth (32-bit float, not 8-bit), otherwise you lose precision in the mid-gray range and get banding artifacts. For very distant objects (sky, horizon), many renderers store a very high distance value — then use logarithmic scaling or intelligent clipping to optimize your working range. Some compositing packages (Nuke, After Effects) have specialized nodes for depth channel operations; this saves you time when fine-tuning depth effects.
A practical workflow: Export your depth channel as a separate EXR layer with your beauty pass. In the edit, you first use it for placing focus planes, then for subtle atmospheric shifts. For VFX shots where you need to insert objects, the depth channel becomes your reference template — this way, you integrate new elements spatially consistently into the existing scene.