Real-time compositing during shoot or color suite — VFX and layers applied live, not pre-rendered offline. Enables instant corrections and tweaks on set.
While you're on set and the first take is in the can, your compositing technician is already sitting at the video village station, assembling the layers live. That's on-line compositing—and it fundamentally changes how quickly you get feedback on VFX elements. Instead of waiting three weeks for color correction and then another three weeks for compositing, you see the final layers during or immediately after the shot. Greenscreen keying, particle effects, 3D inserts—everything is composited in real-time or with minimal latency.
The practice on set is significantly less spectacular than it sounds. You need a dedicated workstation with enough GPU power (usually a Quadro or RTX configuration), specialized software like Nuke, Fusion, or proprietary real-time engines, and, above all, stable data connections between camera capture and the compositing rig. The key advantage: while your DP and crew are still setting up, you can already test on the green background to see if the focus, keying, and lighting are working. If the actor is holding a vase that will later be replaced digitally—you'll know immediately if the position, size, and shading are correct. No surprises in the DCP.
Practical Limits and Workflow
On-line compositing works best for stable, predictable shots: product placements, digital props, keyer corrections, simple tracking jobs. For complex scenes with many moving elements or for shots that still require extensive color correction, it becomes challenging—real-time processing then can't achieve the quality you'll need later. Here, on-line compositing is more of a control tool than a final render.
The workflow typically looks like this: raw footage comes live from the camera, is buffered in RAW format (or ProRes), the compositor sets up their nodes in Nuke—Roto, Keyer, Transform, 3D-Sync—and you get a preview on the monitor. If the quality is usable, the setup is noted and later rendered in post with full resolution and timecode. If something went wrong, you do a second take—but this time with actual feedback, not with hope. This iteratively saves an enormous amount of time in post-production and reduces the number of VFX fixes that arise downstream.