OpenAI expands access to its DALL-E text-to-image model

21 Jul 2022

An AI-generated image from DALL-E. Image: OpenAI

OpenAI is giving beta users free credits each month to generate images from DALL-E, along with full usage rights to commercialise the images they create.

OpenAI is giving 1m people early access to the beta of its advanced text-to-image generator, DALL-E 2.

When users input a phrase or a string of words into this AI system, it is able to interpret the description and create multiple images based on the text prompt.

Under the beta access, users receive 50 credits in the first month and 15 free credits each month after. These can each be used for one original text prompt generation, which creates four images, or an edit or variation to a prompt, creating three images.

Users can also buy credits in the first phase of the beta in batches of 115 for $15. Artists who need financial assistance can apply for subsidised access.

“We are excited to see what people create with DALL-E and look forward to users’ feedback during this beta period,” OpenAI said in a blogpost.

OpenAI said users who create images from DALL-E get full usage rights to commercialise the images, including the right to reprint, sell and merchandise.

The AI research company created the first version of DALL-E last year. The new version DALL-E 2 was unveiled in April, which OpenAI said “generates more realistic and accurate images with four times greater resolution”.

Tackling misuse concerns

The company opted initially to keep its text-to-image generator out of public access in order to properly learn the limitations and safety issues of the AI system.

OpenAI said it has now added new safety features. The system will reject image uploads that contain realistic faces, or attempts to imitate public figures such as celebrities or politicians.

The company said the also system has more accurate content filters to prevent harmful images, such as violent, adult or political content. It has added a new technique to reduce bias in the AI model and “more accurately reflect the diversity of the world’s population”.

“Expanding access is an important part of our deploying AI systems responsibly because it allows us to learn more about real-world use and continue to iterate on our safety systems,” OpenAI said.

Competition appears to be on the horizon for OpenAI, as Google Research revealed its own text-to-image generator called Imagen in May. The Google team behind the model said it had an “unprecedented degree of photorealism” and a deep level of language understanding.

Meta became the latest player to enter the text-to-image AI arena last week, when it revealed its own model called Make-A-Scene. Meta said this system accepts rough sketches from the user to direct the AI before the final image is created.

A publicly accessible text-to-image generator called Dall-E Mini garnered a lot of attention on the internet last month. This AI tool was inspired by OpenAI’s tech and was created by machine learning engineer Boris Dayma.

10 things you need to know direct to your inbox every weekday. Sign up for the Daily Brief, Silicon Republic’s digest of essential sci-tech news.

Leigh Mc Gowran is a journalist with Silicon Republic

editorial@siliconrepublic.com