Task-oriented semantic communication is an important paradigm for machine-type communication in the 6G era, which integrates task-specific goals into system design and optimization. Task-oriented communication bypasses the need to reconstruct the original source, transmitting only task-relevant features to perform the downstream intelligent tasks, which reduces unnecessary data exchange and promotes communication efficiency. This chapter gives an overview of task-oriented semantic communications, with a practical case for comprehension. Moreover, advanced generative models can serve task-oriented communication to achieve higher efficiency.