Framepack AI is an open-source, next-frame prediction video generation model that produces high-quality, 30fps videos up to 120 seconds long. Its unique fixed-length context compression technology minimizes VRAM usage, allowing long video generation on consumer-grade NVIDIA GPUs with just 6GB of VRAM. Developed by the creator of ControlNet and a Stanford professor, it’s optimized for speed, stability, and accessibility across platforms.