An action clip often needs a stable last pose so you can cut to a reaction, place text or continue into another angle. Record enough of the reference after the movement ends. A jerk in the last frames can make an otherwise clean shot unusable.
Define a stopping cue you can see
For a tabletop reveal, record one hand placing a ceramic bowl and withdrawing completely. Hold the frame steady once the bowl is at rest. In the AI video generator, choose a model and mode that accept the reference type. Use a restrained instruction for the bowl position and any required exit direction. Leave a little time after the hand disappears; the placement should not be the final frame.
If a platform takes only part of the reference or duration differs by model, verify the delivered timing. A hold must be confirmed in the output.
Check for secondary movement during the hold
Stop at several evenly spaced points after the hand leaves. Look for the bowl sliding, changing diameter or picking up a new highlight while the surrounding tabletop stays still. If only the final few frames drift, cut at the longest clean hold.
When the clean section is too short for an outro, finish on an approved still or a second static clip. Keep audience reading time in mind for any end text rather than stretching frames that still contain movement.
For an edit that cuts to a different view of the bowl, mark its rim height against the table edge. The held frame is also the best matching reference for the next shot.

