Onsetology Field Notes
Planning AI Post-Production from Camera Through Delivery
A production-minded framework for testing, versioning, compositing, color, and approval across an AI post-production pipeline.
Treat every AI change as a shot task
AI post-production becomes manageable when the team stops discussing it as one broad effect. Each shot has a source, a requested change, elements that must remain stable, a visual reference, a duration, an edit context, and an approval state. That is the same discipline used for compositing and VFX.
A shot audit identifies where the method is likely to work directly, where a conventional VFX step is needed, and where the requested change may introduce continuity or rights concerns. This produces a realistic test plan before the project scales.
Separate generation from integration
A promising generated result is not automatically a finished shot. It may still need tracking, masks, edge work, stabilization, cleanup, grain, color matching, relighting, or selective return to the original plate. The integration plan should be visible in the schedule and review process.
This separation also improves creative decisions. Reviewers can distinguish the quality of the visual idea from the remaining craft needed to place it convincingly inside the film.
Use representative motion tests
The first test should not be the easiest frame. Choose a short section that contains the motion, edges, lighting change, or interaction that makes the sequence difficult. If the workflow survives that test, the team has evidence for scaling it.
Tests should be reviewed in sequence and at delivery speed. Side-by-side comparisons are useful for evaluating whether the requested transformation is actually improving the shot or only making it more novel.
Finish in the same color and delivery context
AI-generated elements need the same color-management and delivery discipline as photographed or conventionally rendered material. Reviews should use known transforms, consistent displays, and the actual edit whenever possible.
Final delivery includes more than the hero master. Aspect ratios, cutdowns, text-safe areas, compression, platform specifications, and archive requirements should be defined before the last round of shot work. That is how an experimental technique becomes a dependable production workflow.
