MotionDirector is a technique that enables the customization of text-to-video diffusion models to generate videos with desired motions. It utilizes a dual-path LoRAs architecture to decouple the learning of appearance and motion, and incorporates a novel biased temporal loss to mitigate the influence of appearance on the temporal training objective. This method supports various downstream applications, such as mixing the appearances and motions of different videos, and adding animations to individual images with custom motions.