OpenAI’s DALL-E took the world by storm earlier this year when it launched its second iteration of its machine learning image generator. While it was initially made available only to a select few, the company is now opening up a beta version of the system to the public, but on a limited basis.
In a blog post, the company announced that it will be inviting one million users from its waitlist within the next few weeks to have access to DALL-E. It did not clarify whether it will choosing the first one million to join or at random, so if you still have not signed up, you can put yourself in the waitlist here.
Each beta user will initially get 50 free credits for their first month and then 15 free credits every month after that. One credit equals to one original prompt generation, which returns four images, or one edit or variation prompt that only gives you three images. If you run out of credits, you can purchase an additional 115 credits for US$15 (~RM67) and for artists lacking in funds, you can also apply for subsidised access.
To expand on that, an original prompt lets you create any fantastical image you can think to write, seemingly out of thin air. Additionally, an edit will allow you to make context-aware changes to images that you have previously created using DALL-E.
A variation, on the other hand, will let you make different variations of the original image, which can either be generated or uploaded by the user. All generations made by DALL-E can be saved by the user in “My Collection”. Moreover, each user gets full usage rights to commercialise the generated images including the right to reprint, sell, and merchandise.
For the beta phase, OpenAI made some changes to DALL-E to improve safety. The AI rejects image uploads with realistic faces and it also prevents the generation of photorealistic faces. Violent, adult, or political content are also blocked from being generated, with both automated and human monitoring systems implemented to prevent misuses.
The post AI Image Generator DALL-E Now Available In Beta appeared first on Lowyat.NET.