SwitchCraft: Training-Free Multi-Event Video Generation with Attention Controls
This addresses a limitation in video generation for users needing coherent multi-event narratives, though it appears incremental as an enhancement to existing diffusion models.
The paper tackles the problem of multi-event video generation where current text-to-video diffusion models produce blended or collapsed scenes, and presents SwitchCraft, a training-free framework that improves prompt alignment, event clarity, and scene consistency compared to existing baselines.
Recent advances in text-to-video diffusion models have enabled high-fidelity and temporally coherent videos synthesis. However, current models are predominantly optimized for single-event generation. When handling multi-event prompts, without explicit temporal grounding, such models often produce blended or collapsed scenes that break the intended narrative. To address this limitation, we present SwitchCraft, a training-free framework for multi-event video generation. Our key insight is that uniform prompt injection across time ignores the correspondence between events and frames. To this end, we introduce Event-Aligned Query Steering (EAQS), which steers frame-level attention to align with relevant event prompts. Furthermore, we propose Auto-Balance Strength Solver (ABSS), which adaptively balances steering strength to preserve temporal consistency and visual fidelity. Extensive experiments demonstrate that SwitchCraft substantially improves prompt alignment, event clarity, and scene consistency compared with existing baselines, offering a simple yet effective solution for multi-event video generation.