DALL-E, as well as DALL-E 2, are artificial intelligence models developed by OpenAI and designed to generate digital images based on natural language descriptions. OpenAI has added a new “outpointing” feature to DALL-E, which allows the system to create new visual effects that push the boundaries of any image, reports The Verge.
DALL-E uses human cues to “imagine” what’s outside the frame of Johannes Vermeer’s portrait Girl with a Pearl Earring. Interestingly, even with limited information, the system can match Vermeer’s style, mimicking the shadows and highlights of the original.
DALL·E: Introducing Outpainting
Extend creativity and tell a bigger story with DALL-E images of any size
blog: https://t.co/dBRzCXWDYF pic.twitter.com/9sOyDMPEb7
— AK (@_akhaliq) August 31, 2022
The time-lapse footage shows how artist August Kamp expanded the image in small sections at a time, often reworking the DALL-E generated images to get the desired result. The system doesn’t generate these extensions by itself: as with any AI that turns text into images, this model requires humans to describe the visual effects.
Outpainting as a function can be used to expand original content, but of course many DALL-E users have been playing with the feature for cartoonish purposes.
From a broader perspective, outpainting doesn’t really expand the basic functionality of text-to-image AI systems, but it does show how OpenAI will likely position itself in the growing market for these systems: by making usability a key pitch to customers.
Many text-to-image AI models can perform the same important function as outpainting, but like DALL-E itself before this update, it requires a lot of manual manipulation. Making outpainting as simple as possible will help DALL-E stand out from the growing competition of smaller systems such as Midjourney and Stable Diffusion.
DALL-E itself is available through a beta testing program, which has so far been accessed by over a million users. Every beta user gets 50 free image generation for the first month, and 15 additional uses every month after that. Over time, 115 additional images can be generated for $15.