Identity Drift Gate

Decide whether a generated VIDEO OR IMAGE SEQUENCE still shows THE SAME CHARACTER across every shot, using THE WORST SHOT as the statistic instead of the average -- and refuse to guess when the gate itself is unconfigured or the calibration data cannot support a threshold. Zero dependencies, pure functions over the standard library.

THE SILENT FAILURE. A sequence of twenty shots where nineteen are perfect and one shows a visibly different person averages to an excellent score. The average is the wrong statistic for this question: a viewer does not experience the mean of a sequence, they experience the shot where the character's face changed, and that shot is exactly the one an average is designed to absorb.

So the verdict is driven by the worst shot. A sequence is only as consistent as its least consistent frame, and that is not a stylistic preference -- it is what the failure actually looks like to whoever watches it.

IT REFUSES RATHER THAN GUESSING in two specific cases: when no threshold has been configured, and when the calibration data provided cannot support one. A gate with a made-up threshold passes everything or fails everything, and either way stops being consulted -- which leaves the sequence unchecked while looking checked.

VERIFIED: 61 tests, measured by running the suite.

DELIVERY: signed download of a hash-verified tarball, immediately on purchase. Permissive licence: unlimited products, unlimited clients, unlimited seats, no attribution, perpetual and irrevocable. One restriction, do not republish the source as source.