A computer-implemented method for rendering video content is disclosed according to an aspect of the subject technology. The method comprises retrieving video content from a network, directing the retrieved video content to a video hardware accelerator, wherein the video hardware accelerator renders the video content into rendered video content, and determining whether the video content is to be displayed in an embedded mode or a full screen mode. The method also comprises, if the video content is to be displayed in the embedded mode, then instructing a graphics processing unit (GPU) to composite the rendered video content with other rendered content for display to a user on a display, and, if the video content is to be displayed in the full screen mode, then instructing the video hardware accelerator to send the rendered video content to the display without compositing by the GPU.