A prediction block is determined for a current block of a current frame of a video stream using a template having pixel locations that conform to a subset of the pixel locations of the current block. A first portion of the prediction block having the same pattern of pixel locations as the template is populated by inter-predicted pixel values, and the remaining portion of the prediction block is populated by intra-predicted pixel values. The intra-predicted pixel values may be determined using inter-predicted pixel values of the first portion, pixel values of pixels adjacent to the current block, or both.