AnimateDiff is an AI tool which generates animated videos from text prompts or static images by predicting motion between frames. It leverages Stable Diffusion models and pre-trained motion modules to create the animation without requiring manual creation of each frame. The process starts with the motion module taking a text prompt and preceding frames as input. It then predicts the scene dynamics and motion to transition smoothly between frames. These predictions are conveyed to Stable Diffusion which generates an image that satisfies the text prompt whilst aligning with the motion predicted by the module. This results in the creation of a smooth, high-quality animation from a textual description. AnimateDiff can be used for a range of applications such as prototyping animations, visualizing concepts, game development, creating motion graphics, animating augmented reality characters, previewing complex scenes, creating educational content, and generating animated social media posts. Users can use AnimateDiff freely on animatediff.org without needing personal computing resources or programming skills. Users simply input a text prompt describing their desired animation and the animation is automatically generated in a short time. System requirements for running AnimateDiff include a powerful Nvidia GPU with abundant VRAM, an operating system that is either Windows or Linux, minimum 16GB of system RAM, and a large quantity of storage (at least 1 TB recommended). AnimteDiff extension is also available for installation with the AUTOMATIC1111 Web UI.