Unlocking Zero-Shot Video Editing with Cross-Attention Guidance in Text-to-Video Diffusion Models

Cross-attention guidance can enable zero-shot control over object shape, position, and movement in text-to-video diffusion models, despite the limitations of current models.