“Visually striking” and “crisper” is how the company described DALL-E 3's intricately detailed images compared to its predecessor. That includes faces (which gen AI is very good at) and hands (which gen AI is shockingly bad at).
The new training model is out for ChatGPT Plus and ChatGPT Enterprise users now, with the AI tool set to be available via the API and in Labs later this year.
Getting down to business
Across an organization, the call is always the same. More content. Better content. Faster content. But mostly more content. So, that’s where a tool like DALL-E 3 comes in, rapidly generating photorealistic images and graphics for use and iteration via text descriptions.
In its big reveal, OpenAI highlighted AI-powered web design and a logo maker as example business use-cases. But really, it’s for anyone who needs a steady flow of branded visual assets. Although whether that’s good or bad is open to debate since the platforms are not without their critics.
DALL-E 3 itself brings a few more upgrades. It’s now better at responding to extensive, detailed prompts to deliver those apparently “striking” images. There’s support for landscape and portrait aspect ratios. Behind the scenes, the model will now be paying closer attention to captions made by users. In other words, it should be better at doing what users tell it to do.
The AI research and development firm also confirmed it’s working on a digital provenance tool. Early tests show a 99% hit-rate when identifying unedited DALL-E generated images. That success rate drops to a respectable 95% when making minor edits to the image, like cropping or compression.
Users can sign up for DALL-E 3 by clicking here. When our Cameras Editor tested it out, he came away calling it the easiest, beginner-friendly AI art generator out there - but found it struggled to produce photorealistic images, warranting the artistic touch of a human hand.
More from TechRadar Pro
Are you a pro? Subscribe to our newsletter
Sign up to the TechRadar Pro newsletter to get all the top news, opinion, features and guidance your business needs to succeed!
Steve is TechRadar Pro’s B2B Editor for Creative & Hardware. He explores the apps and devices for individuals and organizations that thrive on design and innovation. A former journalist at Web User magazine, he's covered software and hardware news, reviews, features, and guides. He's previously worked on content for Microsoft, Sony, and countless SaaS & product design firms. Once upon a time, he wrote commercials and movie trailers. Relentless champion of the Oxford comma.