The relationship between artificial intelligence and art is much broader and more complex than anyone can imagine. We have examples like AI Gahaku's Renaissance portraits, dantesque nudes, and even the reverse effect, creating real faces from paintings and objects. However, creating images with AI using text as the main reference has become incredibly popular recently, which brings us to DALL-E mini, a model that's available to everyone and very easy to use.
Creating art from a phrase isn't exactly new. In December of last year we talked about Wombo.art, which uses that method to generate incredible characters and scenes, but many arrows point to the launch of DALL-E in January 2021, courtesy of OpenAI.
Anyone following OpenAI's activities closely probably already knows that its experts are working on the second version of DALL-E, and they've already shared some very interesting results. However, today we're focusing on one of DALL-E's open source variants, called DALL-E mini. The good news is you don't have to download anything: go to its official page, enter a description, and wait.
How to Create Images with AI Using Text
The DALL-E mini interface offers the bare minimum: a text box and a Run button that starts the image generation. The system is being slammed with requests, so you'll likely get a traffic overflow error at first, but it always works on the third or fourth attempt.
As for the results, the sky is the limit... or rather, the algorithm. Its official page explains that the model is still in training and will improve over time, but users are the real engine. Some are using DALL-E mini to create impossible crossovers, like Columbo in Marvel vs Capcom 2, or strange and disturbing versions of famous characters.
All the relevant code for DALL-E mini is available on its GitHub profile, and it's also worth following the development reports on its official page. As always in these cases, we recommend taking advantage of DALL-E mini while it's online and being patient with its errors.
Official site: Click here
