Vript: A Video Is Worth Thousands of Words

Advancements in multimodal learning, particularly in video understanding and generation, require high-quality video-text datasets for improved model performance.