It uses a custom renderer, which seems to just blit its image data onto a canvas. The HTML is just there so you can actually select the text.