Treat every requested change as a shot task
AI post-production becomes manageable when the team stops describing it as one broad effect. Every shot has a source, a requested change, elements that must remain stable, a visual reference, an edit context, a duration, and an approval state.
A shot audit identifies where an AI method is appropriate, where conventional compositing or cleanup is still required, and where source quality, continuity, reflections, or rights create additional risk. The audit becomes a test plan before work expands across the sequence.
Separate generation from integration
A useful generated result is not automatically a finished shot. It may still need tracking, masks, edge work, stabilization, cleanup, grain, relighting, color matching, or selective restoration from the original plate.
Keeping generation and integration visible as separate stages also improves review. The team can evaluate the creative direction of the change without confusing it with the craft still required to make that change belong inside the photographed film.
Finish inside the real edit and delivery context
Tests should be reviewed in sequence and at delivery speed. A side-by-side comparison can show whether a transformation improves the shot, but the edit reveals pacing, continuity, and whether the new image supports the scene rather than only adding novelty.
Final work follows the same color-management and delivery discipline as photographed or conventionally rendered material. Aspect ratios, cutdowns, text-safe areas, compression, platform specifications, and archive requirements are defined before the last round of shot work.