Modelscope Text to Video Synthesis

ModelScope is a free and open-source text to video generator. The text-to-video generation model has three parts: one that extracts features from the text, another that converts those features into a video, and a third one that turns the video into a visual representation. The model has around 1.7 billion parameters and can handle English input. It uses a structure called Unet3D and creates videos by removing noise from a starting point of random noise.