MemFlow is a real-time optical flow estimation and prediction method that leverages a memory module to store historical motion information. The method aims to address the limitations of classical optical flow estimation, which uses only two frames as input, and multi-frame methods, which often suffer from high computational overhead and require unseen future frames. MemFlow integrates resolution-adaptive re-scaling to handle diverse video resolutions and extends to future prediction based on past observations. It outperforms state-of-the-art methods like VideoFlow and FlowFormer in terms of generalization performance and inference speed on datasets such as Sintel, KITTI-15, and the 1080p Spring dataset. The method's key contributions include innovative real-time optical flow estimation, enhanced generalization with resolution-adaptive re-scaling, superior optical flow estimation, and future prediction capability without explicit training.MemFlow is a real-time optical flow estimation and prediction method that leverages a memory module to store historical motion information. The method aims to address the limitations of classical optical flow estimation, which uses only two frames as input, and multi-frame methods, which often suffer from high computational overhead and require unseen future frames. MemFlow integrates resolution-adaptive re-scaling to handle diverse video resolutions and extends to future prediction based on past observations. It outperforms state-of-the-art methods like VideoFlow and FlowFormer in terms of generalization performance and inference speed on datasets such as Sintel, KITTI-15, and the 1080p Spring dataset. The method's key contributions include innovative real-time optical flow estimation, enhanced generalization with resolution-adaptive re-scaling, superior optical flow estimation, and future prediction capability without explicit training.