In this paper, we propose a planning method for fast autonomous flight. First, the mass-point dynamic and the sample velocity are used to construct the time search graph. To improve the velocity sample efficiency in the entire feasible space, the time-guided method is proposed to refocus on a smaller lattice space. After a global time search based on the heuristic function, trajectory optimization is employed throughout each time interval. We use bézier curves for trajectory optimization, and the time optimal planning is transformed into a convex optimization problem. Compared to other methods, our proposed method has better performance on time and path length.