The talent scheduling problem is a simplified version of the real-world film shooting problem, which aims to determine a shooting sequence so as to minimize the total cost of the actors involved. In this article, we first formulate the problem as an integer linear programming model. Next, we devise a branch-and-bound algorithm to solve the problem. The branch-and-bound algorithm is enhanced by several accelerating techniques, including preprocessing, dominance rules and caching search states. Extensive experiments over two sets of benchmark instances suggest that our algorithm is superior to the current best exact algorithm. Finally, the impacts of different parameter settings are disclosed by some additional experiments.