A device for video coding is configured to determine a characteristic of a predictive block of a current block of a current picture; identify a transform for decoding the current block based on the characteristic; inverse transform coefficients to determine a residual block for the current block; and add the residual block to a predictive block of the current block to decode the current block.