SEAM: Shot Entity-Attribute Memory for Consistent Short-Drama Generation at Scale
2026-08-24 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors address a problem in generating short drama episodes where visual details like props and character positions change inconsistently between shots. They introduce SEAM, a tool that keeps track of important details across shots using a memory graph and fixes these continuity errors by updating the prompt text before generating images. Their system improves consistency significantly as shown on their new SEAM-Bench test and works well with different text models. When used in a professional production setting, SEAM helped achieve a very high acceptance rate from directors by maintaining continuity without unsafe changes.
short-drama generationvisual continuitymemory graphprompt rewritingtext-to-image modelscontinuity recallstoryboardingcross-episode contextCreativeFittingSEAM-Agent
Authors
Jiaqi Liu, Maolin Ran, Xiaoyang Lu, Jian Wang, Weiwen Liu, Jianghao Lin, Yong Yu, Weinan Zhang
Abstract
Short-drama generation has grown into a large, industrialized pipeline, and as it scales from isolated shots to the episode level, visual continuity has become a critical bottleneck. Current agent frameworks generate each shot in isolation, so context drifts across shots and props, character posture, and blocking turn inconsistent. Once assembled, these small discrepancies amplify into severe visual breaks. We present SEAM (Shot Entity-Attribute Memory), a training-free, model-agnostic memory graph that repairs continuity entirely at the prompt-text layer by extracting a multi-dimensional state for every shot, retrieving only causally prior context over the resulting graph, filtering it selectively, and injecting the surviving constraints by natural-language prompt rewriting. We further release SEAM-Bench, a double-blind continuity storyboarding benchmark, on which SEAM raises cross-episode continuity recall from 0.700 to 0.946, generalizes across six mainstream text models, and yields consistent, though not yet significant, gains at the generated-image layer. Deployed as a mandatory stage in CreativeFitting's SEAM-Agent production pipeline over 201 shots, SEAM reaches a 96.5% director-acceptance rate with zero unsafe injections; a conservative counterfactual attributes at least 21.9 percentage points of that rate to its cross-episode memory.