It's been at least a couple of days since an AI image generator announced a major update, so it was about time we had something new to play with. Sure enough, Open AI's DALL-E 3 has added a couple of new features.
DALL-E 3 still remains on one of the best AI art generators despite the advances of rival offerings. It's latest update means that users can now make localised edits to images directly in ChatGPT, and there are new preset style options – a feature of many other basic AI image generators. The reaction is mixed. Some users are describing the update as "amazing" and "epic". Others can't believe the platform didn't already have them.
You can now edit DALL·E images in ChatGPT across web, iOS, and Android. pic.twitter.com/AJvHh5ftKBApril 3, 2024
So the first change is that users can tweak their initial generations in the ChatGPT chat itself, both on the web and on mobile. That includes changing specific details, such as the contents of a cup or the colour of someone's lipstick. The second change is the introduction of style prompts, which appear to be intended to help new users get started if they're not sure what to ask for in their text descriptions.
Created using the new Dalle3 image editing feature in ChatGPT. image to video using pika https://t.co/y3eI3JQVt9 pic.twitter.com/s4wUpfAwI6April 2, 2024
The first feature in particular has been welcomed by many users, but some are surprised that it's taken so long (and also surprised by how long it appears to take to generate changes). After all DALL-E 2 was one of the first AI image generators to add inpainting.
"This is one of the most awaiting features? Could have used Canva for this," one person wrote in response to the announcement on X. "Respect for not editing out how painfully slow that was," someone else commented in reference to the video demonstration. People also want to know if they'll ever be able to edit existing non DALL-E 3 generated images in ChatGPT instead of resorting to Photoshop Generative Fill.
Other recent news in AI image generation is the addition of Structural Reference in Adobe Firefly and the cref tag in Midjourney.
