The AI art generator DALL-E 2 has become something of a household name in a short space of time, and it continues to roll out game-changing features. It's already stunned us with its ability to create photorealistic images of people who don't exist, and now users can do the same with real people. OpenAI, the makers of the software are now letting people use the platform to edit real faces.
The latest development isn't a new feature, but it's one that OpenAI had removed access to because of the fear of misuse. Until now it's only been allowing users to edit human faces created by the AI itself – i.e., people that don't exist. Now it's removed that restriction, and users are going wild putting their own faces on everything from samurai warriors to wrestlers and more... but mainly samurai warriors and wrestlers. Unsure what DALL-E 2 is or how it works? We'll answer some questions below, but see our article on how to use DALL-E 2 for more details.
OpenAI has decided to let users of text-to-image AI art generator and DALL-E edit images of real human faces after having held back the feature to prevent misuse. Like the concerns around deepfake videos, the fear was that people would use the feature to pass off photorealistic images as real for all kinds of nefarious ends. So why allow people to use it now? Well, OpenAI says it's now improved its filters in order to remove images that contain “sexual, political, and violent content.”
It said in an email to users: “Many of you have told us that you miss using DALL-E 2 to dream up outfits and hairstyles on yourselves and edit the backgrounds of family photos. A reconstructive surgeon told us that he'd been using DALL-E 2 to help hos patients visualize results. And filmmakers have told us that they want to be able to edit images of scenes with people to speed up their creative process.
"With improvements in our safety system, DALL·E is now ready to support these delightful and important use cases — while minimizing the potential of harm from deepfakes."
But probably another big influence is the fact that users had been clamouring for the feature to be released and that DALL-E 2's main rivals, such as Stable Diffusion, have been taking a less cautious approach with fewer restrictions, which in turn allows the technology itself to develop quicker.
The terms of use for DALL-E 2 do forbid users from uploading images of people without their consent but that does not seem practically possible to enforce at the moment. Many people had already found ways around the previous restriction, for example by editing faces in sections to avoid them being detected. Users have been celebrating the new freedom, however, mainly by putting their own faces into all kinds of scenes. Here are a few examples from Twitter.
What is DALL-E 2?
DALL-E 2 is a text-based AI image generator. Users type in a prompt defining the image they want to create, including its subject and style, and then the platform creates the image, which can then be fine tuned. The most recent new feature is the addition of DALL-E 2 outpainting, which allows users to expand an image beyond its original frame. People have been using it to "uncrop" masterpieces.
How can I access DALL-E 2?
DALL-E 2 is currently available by invite only and there's a waiting list (I know, not very 'open' of OpenAI, but the company says that more than a million people now have access). The platform is no longer free, but works on a credit-based microtransaction model. Users begin with free credits but then need to top them up in order to generate or edit more images.
AI image generators have been attracting a lot of attention as their output improves, and especially after AI art won a competition. See our roundup of the weirdest AI art for examples of what they're capable of, and take a look at the results of the best AI art generators compared.
Read more:
