Phenaki

Phenaki is an AI-based system that can produce lifelike visuals from text prompts. It relies on a novel causal model to figure out how to represent videos and then condense them into separate tokens for clips of diverse lengths. It is capable of producing videos from open-domain, time-variable stimuli and it performs better than traditional methods of single-frame video generation. Additionally, it requires fewer video-text pairs and larger image-text collections. This particular approach to video generation is a first for the academic field.