The Dall-E AI image generator just entered a limited open beta

Cal Jeffrey

Posts: 3,657   +1,122
Staff member
In a nutshell: If you have ever wanted to create your own art but lack the skills to make anything other than a stick figure, OpenAI's Dall-E might be worth giving a try. The tool is best known for creating unique images that are sometimes surreal or frightening from simple text descriptions. Best of all, the images you create with Dall-E are yours. No attribution or licensing is necessary.

OpenAI's Dall-E image generator is both powerful and weird. So far, the developers have only allowed a handful of AI researchers to use the tool. However, OpenAI announced on Wednesday that it would open the image creator in beta to the public in a limited capacity. The artificial intelligence startup began a waiting list and will soon send out invites to one million people.

Users will receive 50 credits for their first month and 15 additional credits in subsequent months. They can also buy more in 115-point increments for $15. Creators use credits to take a spin on the wild ride of AI image generation. One point produces four unique pictures from a user-generated text prompt. The Dall-E beta will also include Edit and Variation options.

The Edit feature allows users to "make realistic and context-aware edits" to generated images in their collection. They can also use it to make changes to uploaded pictures. So, for example, you could upload a selfie you took on a trip to the beach and have Dall-E add a realistic-looking pair of sunglasses. Using Edit costs one credit and returns three images.

The Variations feature takes a more holistic approach to editing. It can take a Dall-E created or uploaded picture and make three new variations based on the source image. Each use of Variations costs one credit and, like Edit, produces three new creations.

A perhaps more intriguing aspect of the Dall-E beta is that it is entirely free for commercial use. All images users create with the tool are theirs to use how they want. Content creators can use the pictures in blogs or news articles. The art can be printed, sold, or applied to other merchandise like tee-shirts or coffee mugs.

OpenAI says that it has already received interest from people wishing to use Dall-E to create things like "illustrations for children's books, art for newsletters, concept art and characters for games, moodboards for design consulting, and storyboards for movies."

Even here at TechSpot, finding free-to-use photos for articles can be a huge hassle. Depending on how well Dall-E works, you may start seeing AI-generated images in some of our stories. However, some artwork previously created by the machine seemed random and strange, so we'll be interested to see if it can produce images worthy of everyday use on our site.

OpenAI did not indicate when it would start emailing invites, but it will likely begin with a handful of people and gradually build until it reaches one million users. What happens after that point is also a mystery. The developers emphasized that this is a beta phase and that they are looking for user feedback. Based on that, OpenAI will likely tweak Dall-E until they can fully open the tool to everybody.

Permalink to story.

 

p51d007

Posts: 3,368   +3,029
I have people that send me photos they see and say LOOK AT THIS!
Most of the time, I say, that's not a real photo. Sometimes, people don't edit
a photo (photoshopping) correctly. Scale, skew, SHADOWS, lighting are wrong.
But with a lot of "AI" getting better, who knows what things will look like.
It could be misused politically (duh) to crush someone. Some type of watermark
needs to be on photos to prove they haven't been misleadingly altered.
Just the "content aware fill" option in photoshop saves me a ton of time editing photos
where I don't have to clone things to hide background etc on photos I take.
 

Uncle Al

Posts: 9,290   +8,460
I am curious to know if the estate of Salvador Dallie licensed or authorized the use of his images or name (likeness). I checked and it is still an active estate .....
 

Raytrace3D

Posts: 355   +433
I've been part of the alpha and now the beta and it works astonishing well most of the time. I could totally see someone using these to generate illustrations or artwork for a book or stock photos. Their pricing model is a bit strange right now as I'm sure it'll change. 100 credits (queries) for 15 bucks is a bit steep but we'll see. You do get 15 queries a month for free so I guess that's something. During the alpha you'd get 50 credits per day and you can burn through those pretty quickly. I've used over 150 queries in the last day alone. lol
 

Eflow

Posts: 51   +98
I have people that send me photos they see and say LOOK AT THIS!
Most of the time, I say, that's not a real photo. Sometimes, people don't edit
a photo (photoshopping) correctly. Scale, skew, SHADOWS, lighting are wrong.
But with a lot of "AI" getting better, who knows what things will look like.
It could be misused politically (duh) to crush someone. Some type of watermark
needs to be on photos to prove they haven't been misleadingly altered.
Just the "content aware fill" option in photoshop saves me a ton of time editing photos
where I don't have to clone things to hide background etc on photos I take.

Tools like this for sure have the potential to cause harm through deep-fakes. NPR ran a story on Dall-E yesterday and mentioned that it will not accept inputs for specific individuals (e.g. Joe Biden or Donald Trump). That seems like a good restriction, but I'm sure some creative users will find a way around it.