DALL-E enters beta and will invite a million new users – what could go wrong?
If you’ve seen a surreal image of three grinning chihuahuas on a taxi, an astronaut on a white horse, or a fox possibly painted by Monet, you can probably credit DALL-E, OpenAI’s artificial intelligence-powered image generator, which is finally entering beta this week.
Now in version two, DALL-E 2 (opens in new tab) is possibly one of the most talked about AIs since HAL 9000 (opens in new tab). Its images, which are generated based on natural language description fed to it, are realistic, halting, and mind-bending. They’re now also good enough to be featured on the cover of a magazine (opens in new tab).
Up until recently, most people haven’t had direct access to DALL-E and, instead, had to satisfy their AI image urges with DALL-E mini (opens in new tab), a small-scale knock-off built by Craiyon.
Starting today, OpenAI is inviting up to one million new DALL-E 2 users from its waitlist (your author is on it), but not all at once. In a blog post, OpenAI writes that the process will occur “over the coming weeks.”
Pay the AI
There is a small wrinkle (come to think of it, I’d like to feed that phrase into DALL-E). New DALL-E artists will get 50 free credits to create images with 50 different natural language sentences. After they run through the initial 50, OpenAI will provide them with 15 credits per month. That means you could create 15 different surreal images per month. Though, that may feel a bit like the three wishes you get from a Genie: worrying about using a poorly conceived phrase that inspires lackluster DALL-E art and knowing that you’re now down to just 14 more wishes…er phrases for the month.
You can generate more images if you’re willing to pay DALL-E. No, not in the blood and sweat of an aspiring artist, but in cash. $15 will buy 115 credits or 460 DALL-E image creations.
If you get a DALL-E invite, you’ll be entitled to a number of useful features including the ability to use DALL-E to edit images generated through the system or ones you upload. You could, for instance, upload a classic work of art and have DALL-E generate variations based on the art. DALL-E 2 can also seamlessly add new elements to existing images. You tell it what you want to add and where in the image you want to put it. The edits will take into account shadows and reflections to make it look as if the new element was always part of the original art.
Virtual gallery
As for where you’ll store your AI art, OpenAI is offering to keep it all on the DALL-E platform. Maybe eventually you’ll be able to host your own AI art gallery showings.
Even though OpenAI is charging for the creation of images above what you get for free, the non-profit organization does not appear interested in making a hefty DALL-E profit.
In the blog post, the company announced that you retain the rights to any image you generate through DALL-E, including those generated before the platform went into public beta. Now you can take that image of a dog on a bicycle eating an ice cream cone and post it on websites, or put it on T-Shirts and mugs.
OpenAI fully expects DALL-E images to show up on a wide variety of platforms.
“Users have told us that they are planning to use DALL·E images for commercial projects, like illustrations for children’s books, art for newsletters, concept art and characters for games, moodboards for design consulting, and storyboards for movies,” OpenAI wrote in the post.
Safety first
Before you start feeding photos of family, friends, politicians, and celebrities, know that DALL-E will automatically reject any attempt to create realistic pictures of actual people, including celebrities and politicians.
DALL-E 2 beta will also seek to spit out more diverse images “that more accurately reflect the diversity of the world’s population,” notes the post.
There is, of course, no telling what people will feed into DALL-E 2 and what kinds of results it’ll spit out. OpenAI seems committed to maintaining control and steering the platform clear of bias, but if the history of AIs has taught us anything it’s that their ability to learn and adapt to new information makes them unpredictable.
If you’re hoping for a chance to create your own AI art, you can still sign up for DALL-E 2’s waitlist (opens in new tab), though you’ll surely be queued up behind the millions who probably signed up when DALL-E first appeared a few months ago.
Audio player loading… If you’ve seen a surreal image of three grinning chihuahuas on a taxi, an astronaut on a white horse, or a fox possibly painted by Monet, you can probably credit DALL-E, OpenAI’s artificial intelligence-powered image generator, which is finally entering beta this week. Now in version two,…
