Multi-layer image format embedding depth and additional render passes in one file — OpenEXR with Z-depth, ID masks, reflections. VFX pipeline standard.
In everyday compositing, you don't just drop a flat render into the edit – you need control over every damn pixel, and that's where Deep Images come in. It's about a multi-layer image format that allows you to store depth information and multiple render passes in a single OpenEXR file. Instead of three, four separate EXRs for Diffuse, Reflection, Shadow, Z-Depth – everything in one file, hierarchically organized and with depth sorting.
The practice looks like this: your 3D renderer (Mental Ray, Arnold, RenderMan) doesn't just output RGB on the final render, but also the Z-channel (depth value of each pixel), Object IDs, Material IDs, Cryptomatte masks, and any number of additional render layers like Emission, Ambient Occlusion, or Subsurface. All that stuff lands in a structured EXR, which the compositor later unpacks in Nuke or After Effects. The key: with Deep Images, you can perform depth sorting within a pixel – if multiple objects are layered on top of each other, the depth values are preserved, and you can manipulate them later without producing ordering errors.
On set or in the render lab, you no longer ask: "Can we render another shadow pass?" – no, it's already in the Deep Image. You have your Z-Depth, your ID masks are there, your reflections are separated. This significantly speeds up the compositing process because the compositor doesn't have to wait for ten different output files but receives a single, fully equipped file. This saves storage space (relatively speaking) and prevents errors due to faulty or missing passes.
Important: Deep Images are not the same as Deep Compositing (which works with point clouds and depth samples), but a format standard. Not every software can open Deep Images out of the box – this is a professional feature you'll find in high-end VFX pipelines. Smaller studios often still render in flat layers. But if you need to work with motion blur, complex transparencies, or multiple overlapping effects, there's no way around Deep Images.