Testing AI character consistency across a 20-image travel sequence

SiennaVaire · reddit · 2026-09-23

The author is building a recurring fictional AI character, Sienna Vaire, and stress-testing identity preservation over a 20-image "Leaving London" set: same outfit, luggage, station and weather, while changing pose, framing and action.

The four published frames focus on three checks: face consistency, clothing continuity, and whether the sequence reads as one continuous scene rather than separate generations — the core difficulty of multi-image character generation.

Original post →

More from Multimodal

Multimodal channel →