VBench provides standardized benchmarks and metrics for evaluating text-to-video and image-to-video generation models. The framework includes a curated dataset of evaluation prompts, automated assessment pipelines covering aspects like temporal consistency, motion quality, and text-video alignment, and a public leaderboard hosted on HuggingFace. It supports evaluating models such as those built on Stable Diffusion and serves as a reference standard for the video generation research community.