DOG: Dream Of Galaxy patches together colorful, AI-generated moving images to tell a touching story of a dog cadet selected to travel to space. The turning point of this dog’s life is when an anonymous man asked the stray dog, “Do you want to dream a different dream?”

The main AI technologies used are Stable Diffusion, Midjourney, EbSynth, ChatGPT, etc. ‍

‍

Process

The film was completed in the spring of 2023. At that time, there were no AIs such as Runway, OpenAI Sora, and Pika that could generate videos directly from text. The AI technologies we mainly use include Stable Diffusion, Midjourney, EbSynth, ChatGPT, etc. Because it is a Hackathon competition, the production time is only almost one day. In the early stage of our technical process, we mainly determine the script prompt words through the script. In terms of vision, Midjourney is mainly used to generate color images, and then customized Stable Diffusion is used to generate black and white depth maps based on the color images. Then the color images and depth maps are combined in Cinema 4D and then stretched and extruded to realize a 2.5D model. By rearranging movement, focus, depth of field, etc. for each shot, and then rendering and exporting, a controllable movie language that serves the narrative is realized.