Reproduce “VIRES: Video Instance Repainting via Sketch and Text Guided Generation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Shuchen Weng, Haojie Zheng, Peixuan Zhang, Yuchen Hong, Han Jiang, Si Li, Boxin Shi
A framework called VIRES enables precise, instance-level video editing by integrating sketch and text guidance into a Diffusion Transformer (DiT) model. It consistently achieves superior visual quality, temporal consistency, and adherence to user input, further supported by the introduction of the VIRESET dataset for training and evaluation.
Key insights
- Building upon pre-trained text-to-video Diffusion Transformers provides a robust foundation for temporal consistency in controllable video editing.
- Combining structural sketch guidance with textual appearance descriptions enables precise, fine-grained control over individual video instances.
- Specialized architectural modules and a dedicated instance-annotated dataset are crucial for effectively integrating multimodal guidance and advancing video editing capabilities.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.