Emu Video is a simple text-to-video generation method based on diffusion models, which decomposes the generation process into two steps: first, generate images based on text prompts, and then generate videos based on the prompts and generated images. The decomposition-based generation method can efficiently train high-quality video generation models. Compared with previous methods, our method only uses two diffusion models to generate videos with a resolution of 512 pixels, a playback speed of 16 frames per second, and a duration of 4 seconds.