AI summaryⓘ
The authors address the challenge of keeping characters, objects, and locations consistent across multiple shots in video generation, which is important for making coherent stories. They created EntityBench, a new dataset with many episodes and detailed tracking of entities across shots, along with an evaluation system that measures quality, how well the generated content follows prompts, and consistency across shots. They also developed EntityMem, a method that uses memory to remember visual details of each entity before generating video. Their tests show that remembering entities explicitly helps keep characters looking right better than other methods, especially when the same characters reappear after many shots.
multi-shot video generationentity consistencyvisual narrativesbenchmark datasetprompt alignmentmemory-augmented generationcross-shot evaluationentity trackingvideo synthesisnarrative coherence
Authors
Ruozhen He, Meng Wei, Ziyan Yang, Vicente Ordonez
Abstract
Multi-shot video generation extends single-shot generation to coherent visual narratives, yet maintaining consistent characters, objects, and locations across shots remains a challenge over long sequences. Existing evaluations typically use independently generated prompt sets with limited entity coverage and simple consistency metrics, making standardized comparison difficult. We introduce EntityBench, a benchmark of 140 episodes (2,491 shots) derived from real narrative media, with explicit per-shot entity schedules tracking characters, objects, and locations simultaneously across easy / medium / hard tiers of up to 50 shots, 13 cross-shot characters, 8 cross-shot locations, 22 cross-shot objects, and recurrence gaps spanning up to 48 shots. It is paired with a three-pillar evaluation suite that disentangles intra-shot quality, prompt-following alignment, and cross-shot consistency, with a fidelity gate that admits only accurate entity appearances into cross-shot scoring. As a baseline, we propose EntityMem, a memory-augmented generation system that stores verified per-entity visual references in a persistent memory bank before generation begins. Experiments show that cross-shot entity consistency degrades sharply with recurrence distance in existing methods, and that explicit per-entity memory yields the highest character fidelity (Cohen's d = +2.33) and presence among methods evaluated. Code and data are available at https://github.com/Catherine-R-He/EntityBench/.