A virtual image display device including an image generator for generating a picture light beam modulated based on a video signal, and an optical element including a plane of incidence to which the picture light beam emitted from the image generator is input, and an exit surface emitting the picture light beam after a cross-sectional area of the picture light beam input to the plane of incidence has been enlarged. The optical element includes a first light guide and a second light guide connecting the plane of incidence and the exit surface and guiding the picture light beam. A first light branching layer is disposed between the first and second light guides, and partially reflects and partially transmits the picture light beam. The picture light beam emitted from the image generator obliquely enters the first light branching layer.