Character drift is the gradual change in a character’s face, costume, body, voice, or performance across generated shots. It happens when each prompt or model call treats the character as a fresh request instead of inheriting a locked identity. Even small wording changes between prompts can shift facial structure, hairstyle, wardrobe, or attitude, so a sequence assembled from separate generations stops feeling like one continuous performance.
See how GATA eliminates this by structure rather than by prompt discipline: character consistency in GATA.
What it looks like in practice
A 40-second ad has eight generated shots of the same protagonist. Shots 1–3 look like the same person. Shot 4 was regenerated to fix a hand pose and now has a slightly squarer jawline. Shot 6 inherited a 'closeup, warm lighting' prompt and the hair colour shifted half a tone. By the rough cut, the character feels like four near-identical actors playing the same role. Reviewers notice instantly even if they can't articulate why.
Why it matters
Drift makes a scene feel disconnected and forces teams to spend extra time regenerating, rewriting, or hiding shots that should have matched. A structured workspace avoids it by attaching a single locked character record to every downstream shot, so identity, costume, and tone are inherited rather than re-described each time.
How to prevent it
Stop describing the character in every prompt. Lock the cast once — a single approved face, costume, and performance reference — and attach that record to every shot. Each new generation then inherits the locked identity instead of negotiating it from scratch. In GATA, locking a cast member means downstream shots reference the record rather than re-describing the person, so identity is fixed by structure, not by prompt discipline. Regenerations stay on-model because the source of truth is the cast record, not the latest prompt string.