The dawn of AI is upon us. As DALL-E users experiment with what images a text prompt can produce, Meta has recently unveiled a video AI generator that has taken the internet by the storm.
The results have a nightmarish, dream-like quality to them that is often disturbing. Researchers, however, note that the ‘Make-A-Video’ model “that has only seen text describing images is surprisingly effective at generating short videos.” The model can also use images and other still videos to create variants, much in the same way that it uses text prompts.
The AI uses a diffusion technique for creating images that works backwards from a visual static, but is different from its image counterpart in that it is given “unsupervised training”, meaning it examines the data itself with no guidance from humans.
The researchers continue: “In all aspects, spatial and temporal resolution, faithfulness to text, and quality, Make-A-Video sets the new state-of-the-art in text-to-video generation, as determined by both qualitative and quantitative measures.”
The model isn’t available to the public yet, but interested users can sign up to get on the list.
If you see something out of place or would like to contribute to this story, check out our Ethics and Policy section.