The aerodynamic design of tiltrotor requires a comprehensive consideration of various states such as hovering, cruising, maneuvering, and transition, and higher requirements are placed on the design method. This paper presents a fast prediction method for surface pressure distribution of tiltrotor airfoils based on variational autoencoders. The airfoil shape parameters and corresponding pressure distribution are used as high-dimensional data to train the variational autoencoder model. The model maps high-dimensional data to a low dimensional hidden space and extracts the low dimensional feature structure of the high-dimensional data. The research results on pressure distribution prediction indicate that variational autoencoders can achieve high-precision prediction of pressure distribution and corresponding airfoil shapes. This article develops a hybrid inverse design method for airfoils, which aims to comprehensively consider the characteristics of pressure distribution and the drag performance, use global optimization algorithms to optimize in low dimensional hidden space, and return to the design space to obtain the aerodynamic shape and pressure distribution at the same time. The given target pressure distribution eliminates the strong dependence on designer experience, achieving rapid inverse design of airfoil pressure distribution and achieving good design results, further verifying the prediction accuracy and adaptability of variational autoencoder.