AnimateDiff

AnimateDiff is an open-source framework for generating short, high-quality animations from text prompts, reference images, or existing video using diffusion models. Built around Stable Diffusion and related architectures, it extends image generation into the temporal domain by adding motion-aware components that preserve consistency across frames while allowing flexible control over movement and style. Users can create looping animations, character motions, and stylized sequences by defining keyframes, prompt schedules, and motion strength parameters.

The tool supports multiple model backends, including SD 1.5 and SDXL-based variants, and is compatible with popular UIs such as Automatic1111 and ComfyUI, making it accessible to both researchers and creative practitioners. AnimateDiff offers features such as motion modules, per-frame prompt control, image-to-video pipelines, and integration with ControlNet for pose, depth, or edge-guided animation.