This work introduces a framework that utilizes text-to-image diffusion models for text-driven video editing, improving visual quality and user control significantly.
The proposed method generates high-quality videos that adhere to target text while preserving spatial layout and motion, demonstrating high consistency through diffusion feature space propagation.
Our approach leverages inter-frame correspondences within the diffusion feature space, requiring no additional training and proving effective alongside existing text-to-image editing techniques.
Despite advancements in text-to-video models, challenges remain in resolution and complexity; our framework demonstrates state-of-the-art results addressing these limitations.
Collection
[
|
...
]