Trialled by Daniel Zuboff
Strategy at Loomery
Editing videos can be a very time consuming process. Just rotoscoping a foreground element from a green screen can become laborious, let alone removing someone in the background of a video, or rendering a fully computer generated clip. Fortunately, relatively recent advances in AI have enabled much of this to be done with a few clicks.
Runway is a website that unifies many of these features in one, easy to use place. Below is just a selection of tools, but Runway is able to vary images in ways that you specify, colour grade with just a text prompt, and colourise black and white images and clips. It also offers a bunch of audio features like removing unwanted background noise, or adding subtitles.
Let’s try out a few of these features. How about single image generation?
Prompt: An exciting, modern, intuitive user interface for an app for university students
Okay, this is interesting. It clearly understands some of the features of a phones UI, like the aspect ratio, the icons at the top, and the home bar on an iPhone. Because of the way diffusion-based models work, they’re good with detail, but get confused with global features. You can see this in how the multiple screens the AI chose to generate are merged with each other.
As for the UI itself, there’s not enough novelty or detail to inspire a designer. The two screens in the middle are this odd green colour with little variation, the leftmost UI is pretty basic and the rightmost is a mishmash of boxy shapes with different colours. Lots of image generators struggle with text and other details so let's see how it fares with something a bit more organic.
Prompt: A parrot dressed in a suit working on their laptop in a Parisian cafe with the Eiffel Tower in the background