Clean up grainy low-light footage, high-ISO clips, and noisy phone video. Where the picture is still, each frame is averaged with the three frames before it, which removes random noise while keeping texture. Where something moves, the frame is smoothed on its own instead.
Why video denoising needs more than one frame
Sensor noise is different in every frame; the scene mostly is not. A pixel on a wall has the same true value in consecutive frames plus fresh noise each time, so averaging four of them cuts the noise in half without touching the wall’s texture. A single-frame denoiser has no second sample and can only blur, which takes fine detail with it.
Motion and ghosting
Averaging a moving subject across frames would leave trails. Before blending, each earlier frame is compared with the current one over a small neighbourhood around every pixel. Where they match, the earlier frame counts fully; where they differ by more than noise explains, it counts less, and not at all for clear motion. The noise threshold rises with Noise Reduction, so stronger settings clean more but tolerate smaller movements as noise.
Controls
Noise Reduction sets the overall strength of both passes. Noise Scale sets the size of the grain the spatial pass targets, as a share of the frame. Edge Protection keeps outlines, text, and hair out of the spatial smoothing, and Color Noise sets how hard the colored speckle of low-light footage is suppressed.
Tripod and handheld footage
Locked-off shots benefit most, since almost everything stays still between frames. Handheld footage moves every pixel slightly, so more of the frame falls back to the spatial pass; the result is still cleaner than a single-frame denoise, but by less. Follow with the sharpen video tool at a low amount if the result looks soft.