Text Based image generation accelerates reading

People are incredibly fast in getting the gist of a complex scene. In light of this, repeated text-based image generation using written language (e.g. books, articles, papers) can greatly increase reading speed. Such a system would turn paragraphs into movie-like scene sequences. This joins visual mnemonic mining in applications of generative models in learning. What’s more this approach would leverage the benefits of learning multi-modal representations.