Python · Pyblish · ShotGrid
Playblast and review pipeline
Dailies depend on playblasts looking the same from every artist. This system made capture consistent and automatic: the right cameras, overlays, colour and resolution every time, followed by a queue of steps that encoded, published and opened the result for review without anyone touching it.
How it fits together
CollectShot, cameras, frame range and settings
ValidateChecks with one-click fixes
CapturePer camera, witness cams or full sequence
Post actionsEncode, comp, publish and push to review
What I built
Capture
- Per-camera, witness camera and full sequence playblasts from one tool.
- HUD overlays with shot, frame, focal length and artist details.
- Colour-managed output and a resolution resolver that follows project and shot settings.
- A headless mode so the same capture could run on the farm without a UI.
Publish framework
- Around two dozen collect, validate, extract and integrate plug-ins.
- Validators with fix and select actions so artists could resolve problems in place.
- A metadata sidecar written next to every capture for tracking and later tools.
Post action queue
- A configurable queue of steps driven by YAML, so shows could add or remove steps without code.
- Steps for encoding, frame counters, metadata burn-in, managed storage and registration.
- Multiview and stereo comps rendered through a Nuke template.
- Versions created in ShotGrid with status updates, then opened in RV.
Artist experience
- A tool UI that remembers each artist's settings between sessions.
- Clear progress and error reporting, so a failed step pointed at the cause.
Scale
~14 post actionsconfigurable per show
~25 publish plug-inscollect, validate, extract, integrate
Headlesssame pipeline in the UI and on the farm
Many releaseswith a dedicated test suite
This was studio work, so the code stays with the studio. This page describes what the system did, not how it was written.