CorridorKey 2.0: neural keying inside familiar editing tools

CorridorKey 2.0 is Corridor Digital’s neural keying tool for separating a foreground subject from a screen background. Its official page describes local processing inside editing and compositing applications, with a standalone app also included in the purchase.

Work on difficult edges and transparency

The product says it predicts foreground colour in semi-transparent areas, aiming to remove screen colour from hair, motion blur and other partially transparent details. Its time-awareness option analyses a video clip across frames to reduce flicker and chattering mattes.

These are vendor descriptions of the intended behaviour. A useful test would include the difficult material in a real project: fine hair, a moving hand, reflected screen colour and compression noise. Review the composite over both light and dark backgrounds, rather than judging the result only against a checkerboard.

Keep the artist’s intent in the process

CorridorKey accepts AlphaHint inputs to indicate what should be included or excluded. The page describes supplying one rough guide frame or a guide video for additional control.

That can matter where a colour-based selection is ambiguous or the scene contains equipment an artist wants removed. Start with the automatic result, inspect the affected region and add a hint only when the intended boundary needs clarification. Compare the whole clip after a change, including frames where the subject moves quickly.

Confirm the host, hardware and access

The site lists integrations for Resolve, Nuke, After Effects, Premiere and Blender, alongside a one-time purchase model. Its detailed requirements distinguish Windows systems with NVIDIA hardware from supported Apple-silicon Macs; Intel Macs, AMD graphics and Linux are not currently supported in that requirements section.

The page also carries beta, launch and regional-availability notices. Confirm the current version, host support and checkout availability before buying. The advertised perfect-key and production-use claims are not Franklin benchmarks. Evaluate one representative shot on the intended machine, including processing time, temporal consistency and the amount of manual cleanup still required.