This is the repo for the Video-LLaMA project, which is working on empowering large language models with video and audio understanding capabilities. Video-LLaMA is built on top of BLIP-2 and MiniGPT-4.