What is Optical Flow?

How it works

Optical flow computes the apparent motion of pixels between consecutive video frames, producing a vector field that describes how each pixel moved. In AI video, optical flow is used internally by many models to maintain temporal consistency, ensuring objects move smoothly rather than jumping between positions. It is also used in post-production for frame interpolation, motion tracking, and stabilization. Understanding optical flow helps explain why some AI-generated videos have smooth motion while others exhibit temporal artifacts: models with better optical flow estimation produce more physically plausible movement.