A method of video processing includes performing a conversion between a video comprising a video picture that includes multiple sub-pictures and multiple video blocks, and a coded representation of the video according to a rule. The rule specifies that a boundary between any two sub-pictures is also a boundary between two video blocks.