Picture lock is the moment the edit is declared final so that finishing work can begin: sound design, music, colour, and graphics are all built against a timeline that will not move.
The commitment is the value. Every finishing discipline works to specific frames, so a two-frame trim after lock does not cost two frames of work, it costs a re-conform of everything downstream.
What lock commits you to
Duration, frame-exact. Music was cut to it, sound effects were placed against it.
Shot order and selection. Colour was graded shot by shot.
Titles and timings. Graphics were animated to specific in and out points.
Anything that changes after lock ripples through all four.
Why it gets broken and what that costs
It gets broken because someone senior sees the piece for the first time late, or because a stakeholder who was not in the review chain appears with an opinion. The cost is rarely the edit change itself. It is the re-conform: sound re-synced, graphics re-timed, colour re-linked, and a delivery date that moves.
The preventions are unglamorous and they work. Everyone who can say no attends the fine-cut review. Lock is announced explicitly, in writing, with the consequences named. And a genuine change after lock is treated as a scoped change order rather than a favour, because that is what it costs.
How to lock properly
Announce it. "We are locking picture on Thursday; changes after that affect the delivery date." Ambiguity here is what produces the Friday afternoon trim request.
Get sign-off from the decider, not consensus from the room.
Export a reference with burnt-in timecode and distribute it, so any later note is unambiguous about where it applies.
Freeze the source material too. In generated work this matters more than it used to: if a shot came from a generation, the exact output that made the cut needs to be the one everyone refers to, since regenerating produces something similar and not identical.
The AI-era wrinkle
Cheap regeneration puts real pressure on lock. When a replacement shot costs 54 to 448 credits and twenty minutes, roughly $0.54 to $4.48 at pack rate, "let's just try another version" is always available, and lock becomes a discipline rather than a physical constraint.
The teams that handle this well move the pressure earlier: they generate variants aggressively before lock and then hold the line hard afterwards. The teams that handle it badly discover that infinite cheap options means no piece is ever finished.
Related concepts
What is a rough cut covers the stage where structure gets proven. What is creative review covers the review process that should end at lock. Client approval workflow for AI creative covers getting the approvals in the right order.
Lock holds better when everyone reviewed the same versions in the same place. The 8frame canvas is free and unlimited, and generation is paid from $19/month.