OpenAI announced its "Outpainting" function added to the text-to-image artificial intelligence (AI) model DALL-E, which enables the system to generate new visuals of an existing image that expands its borders.
DALL-E is one of the most popular text-to-image generators, with over a million users. It generates nine 1,024-pixel by 1,024-pixel images that illustrates the command and uses the diffusion process. Until now, it has a limited size and aspect ratio, and anything larger is out of the question. But the outpainting feature generates new visuals of what could be the outside of the frame.
A man films the masterpiece "Girl with a Pearl Earring" by Dutch painter Johannes Vermeer during a preview of the exhibition "The Myth of the Golden Age, From Rembrandt to Vermeer" at Palazzo Fava in Bologna on January 30, 2014.
Outpainting Expands Original Images
One of the examples DALL-E's outpainting feature was used is on the frame of Johannes Vermeer's portrait "Girl with a Pearl Earring." The Verge reported that even with the limited information about the portrait, the system matched the artist's style and mimicked the shadows while still highlighting the original image.
A timelapse video showed how August Kamp used outpainting to expand the image in small sections at a time and even redo some of DALL-E's generations to get the outcome she wanted. However, it should be noted that the system is not generating these extensions by itself because it still needs humans to describe the new visuals.
Outpainting is a feature in DALL-E that expands the original content to show what could lie outside the frame of famous images or paintings. Although it does not expand the basic functions of AI, it does show how OpenAI will compete in the growing market of these systems by making usability its focus for customers.
Making outpainting as easy as possible will help DALL-E stand out from its competitors, such as Midjourney and Stable Diffusion. DALL--E is currently available via a beta program with more than a million users, wherein each could get 50 free image generations in their first month.
ALSO READ: OpenAI Text-to-Image Generator 'DALL-E 2' Now Edits and Combines Existing Pictures
Is DALL-E the AI Artist, A Threat to Human Artists?
DALL-E combines the names of the surrealist artist Salvador Dali and the lovable garbage robot in Pixar-s WALL-E. The first artworks it released were nothing more than curious novelty and were only regarded as remarkable because AI generated them.
Then in January 2022, its new art revealed that it is one of the most advanced image generators in development and is advancing at an astonishing speed. Then recently, OpenAI released the Outpainting feature that allows users to extend their canvass beyond the original frame.
Megan Paetzhold, an artist, wrote in the Intelligencer about her experience using DALL-E and discussed whether it could someday take over people's jobs, particularly the artists. As a photo editor who finds or produces visual elements accompanying her articles in New York Magazine, she wants to see how well DALL-E could do her job.
In her test, she and DALL-E received prompts that told them what to make. They synthesize and produce visuals that are compelling and accurate. At the end of the test, they both produced great art, but Paetzhold explains that DALL-E would not be able to generate the art if it were not for human guidance and knowledge.
That means creating creative output using DALL-E still needs human creativity and knowledge to know how to describe those ideas. For now, the ability of DALL-E is dependent on the imagination of the user behind it, but Paetzhold noted that the advancements of AI are so rapid that it is hard to shake the feeling of a major paradigm shift in how jobs are done.
RELATED ARTICLE: Is AI Art Really Art? Technical Progress in Computer Vision Led to More Self-Described AI Artists in the Past Few Years
Check Out More News and Information on Artificial Intelligence in Science Times