A video decoder configured to determine a block of video data is intra predicted using an angular intra prediction mode, wherein the angular intra prediction mode is one of a bottom-left intra prediction mode or a top-right intra prediction mode; determine an aspect ratio of the block; locate one or more reference samples corresponding to the angular intra prediction mode; apply position dependent intra prediction combination to the reference samples to determine modified reference samples based on the aspect ratio of the block; and generate a predictive block for the block based on the modified reference samples.