If you’ve been hanging out on Twitter lately, you’ve probably noticed a profusion of AI-generated images sprouting all over your timeline, like weird algorithmic visions. These images were generated using a new app called Dream, which allows anyone to create âAI-powered paintingsâ by simply typing in a brief description of what they want to see. It’s weird stuff, often weird – and extremely funny.
The resulting artwork has its own particular aesthetic, defined by swirling shapes and inconsistent objects. The real magic, however, is that no matter what you type, the app will generate something that is visually compelling (at least until we get too used to these toys) and that often matches your prompts surprisingly. appropriate.
Consider, for example, the image below: âGalactic archeology with stars poor in metalâ. Not only has the app created an image that captures the stunning galactic scale of a nebula, but the starry reflections that dot the space are mostly blue – a scientifically accurate shade for poor in metal stars (because the metallicity affects their color).
Some quickly done research on Twitter reveal many more examples, but really, you should play around with the app yourself to understand it better. (If nothing else, the images it generates are exactly the right size to create a custom wallpaper for your phone.)
This kind of AI-generated artwork isn’t new, but it’s getting better and more accessible. Earlier examples of these types of text-to-image models have included research-oriented programs like DALL-E and VQGAN + CLIP, as well as more specialized commercial projects like Artbreeder (which is particularly effective for creating portraits of beings. and fictitious people). . With such tools, the AI ââart scene has exploded in recent years, with practitioners creating everything from lifelike Roman Emperors to infinite waifus.
The Dream app goes even further with its speed, quality and accessibility. It’s available on iOS, Android, and the web and is the work of a Canadian startup named Wombo. The company previously created this AI-powered app that lets you feed static images to create lip-synced renditions of memorable songs. It’s unclear what exactly powers Dream (we reached out to Wombo to find out), but much of the technology in the art of AI is open source, which means the company has likely gone. built on previous work to build the app.
Typically, programs like these are trained on vision datasets – huge libraries of images that are labeled according to objects and landscapes. The programs pick consistent patterns and themes from these images, and then use that information to try to generate something that matches the users’ prompt. We don’t know what dataset Dream’s algorithms were trained on, but based on its output it’s safe to say that it includes a wide range of images – capable of generating images that match. anime characters and video games.
Found an app that is an AI trying to make art and honestly ??? This shit breaks out. I could never have made these colors so vibrant.
These were all created with the words âKnight Armorâ in Wombo Dream. Like??? They are wonderful??? pic.twitter.com/mEANARv8Qm
– MotherLyra (@ Lyraa121) November 21, 2021
The accessibility of Dream means that it is being used for new uses as well. It was used for viral games (like typing your Title of the doctoral thesis and sharing of the result) as well as for more oriented projects. In an amazing Twitter thread, writer and illustrator Ursula Vernon (who posts as T. Kingfisher) shared a short comic they had done using Dream. The comic book characters are hand-drawn, but the backgrounds are AI-generated, with the surreal and shifting quality of the images explained due to the setting: a library of dreams overseen by the Egyptian god of the writing, Thoth.
Vernon tweeted about her experience, noting that she had to do a fair amount of work to prepare the images and that the program’s inability to create landscapes from a space with cohesive architecture created her own challenges.
âIn conclusion, does it work visually? I think the answer is “sort of”. Vernon tweeted. âI’m very aware of the quirks as an artist, obviously. As a dream streak, the messed up architecture does work for a bit, but how long can you get by? Sooner or later the reader will likely notice that nothing happens in the same scene from a different perspective.
So this weekend, armed with a few AI art programs, I started digging around to see what I could do, and if I could put together any of my Weird Little Comic ideas using mostly computer generated images. retouched.
These nine pages were the result. pic.twitter.com/POXoBN0Hbx
– Kingfisher and wombat (@UrsulaV) December 6, 2021
Despite its obvious limitations, Dream gives us a glimpse into the future of synthetic or AI-generated media. For evangelists in this space, the promise of technology is infinite variety. In the future, they say, games, comics, movies, and books will all be generated on the fly in response to our every prompt and whim. And although we are long, long A far cry from such media corresponding to the quality of human production, limited hybrid applications will arrive sooner than you might think – appearing like something first glimpsed in a dream.