CorridorKey 2.0

Overview

CorridorKey 2.0 is Corridor Digital's first official plugin release, a chroma keying tool for green and blue screens. It runs locally in supported VFX and editing software, and a standalone app is included for bulk shots.

Key Features

• One-click keying for most shots, with the plug-in distinguishing background from subject.
• Time awareness analyzes the whole video clip to reduce flicker and chattering mattes and to preserve fine details such as hair.
• Predicts the color of semi-transparent foreground objects, removing screen color for motion blur, bokeh, and hair.
• AlphaHint inputs let artists guide what to include or exclude, including tracking markers, screen edges, and c-stands.
• Supports native 4K and higher, HDR, floating point color depth, any color background, and local processing.
• Plugins for DaVinci Resolve, Nuke, After Effects, Premiere Pro, and Blender, plus an included standalone app.

Use Cases

• Keying green and blue screen shots in Resolve, Nuke, After Effects, Premiere Pro, or Blender.
• Processing bulk shots with the included standalone app.
• Preserving motion blur, bokeh, and hair on semi-transparent or detailed foregrounds.
• Using alpha hints to handle tracking markers, screen edges, and c-stands without garbage masking.

Getting Started

• Set up an NVIDIA GPU environment with ffmpeg on PATH and CUDA-enabled PyTorch, then download the model weights with python scripts/download_weights.py.
• Use the drag-and-drop launcher, or run python run_interactive.py /path/to/clip.mp4 for a guided session.
• Run python infer.py --input /path/to/input.mp4 --output_dir /path/to/output for native-resolution inference with no external hint.

Deployment & Requirements

• NVIDIA RTX 30 series or newer is recommended; RTX 20 series is beta and still being validated.
• 4 GB of GPU memory is required for shots up to 1440p, and 6 GB for 4K.
• Windows 11 or Windows 10 22H2 is supported; Mac beta requires M-series hardware and macOS 14 or newer.
• Repository setup requires ffmpeg on PATH, a Python environment with CUDA-enabled PyTorch, and about 5 GB of model weights; AMD GPUs and Linux are planned.

Before You Adopt

• License: Other. Review its terms before using, modifying, or distributing the project.