A computer implemented method and apparatus for editing video scenes based on learned user preferences. The method comprises identifying a plurality of scenes in a video; for each scene of the plurality of scenes: applying an effect to a representative frame of a scene, wherein the effect is based on learned user preferences; receiving an input that adjusts at least one effect on the representative frame; updating user preferences based on received input; and applying the at least one effect to all frames of the scene; interpolating the adjusted at least one effect at a boundary of two consecutive scenes; and applying the interpolated at least one effect to transitional frames between the two consecutive scenes.