Currently, the most sought-after application of AI video technology is text-to-video AI. The idea is anyone will be able to type a prompt into a text field, and within seconds — or, more likely, minutes — a video will be created using an AI model. Some of the biggest companies in tech are researching this. Meta announced Make-A-Video, Google has announced Imagen Video and Phenaki, and Nvidia says it’s getting involved, too.

All of those projects aren’t available for public use, and the output of these systems is far from photorealistic: figures warp and morph, and there are obvious changes between frames. But the demos are public statements of research and development in the text-to-video field from large — and powerful — companies.

As a creator, the idea of my job so quickly being replaced by AI is terrifying, but as a tech nerd, I am intrigued. I wanted to see how close we are to being able to simply type text into a website and receive polished video. So in the latest episode of my YouTube series Full Frame, I tested five AI video creation tools — Gen-2, Kaiber, Deep Nostalgia, Synthesia, and Unboring — to see what the limits of our AI tech currently are.