Meta unveils Movie Gen, a new model that creates realistic videos with sound from text prompts

Meta has introduced a new generative video model called Movie Gen, which creates realistic videos with sound from text prompts. Unlike previous models, it includes audio that matches the video content, such as engine noises or background music. Movie Gen allows for simple text-based editing, enabling users to make specific changes without altering the entire video. It generates videos at 768 pixels wide, upscaled to 1080p, and can produce clips of up to 16 seconds at 16 frames per second. Currently, Movie Gen does not include voice generation, likely due to technical challenges and concerns about misuse. Meta has stated that this model is for research purposes only and will not be publicly released.