Digital arts are always searching for new prompts to generate the next big thing in AI art. A new feature on the generative AI design site Midjourney—called Describe—lets creators upload a picture as the basis for their art projects. Launched this week, it aims to take the guesswork out of generating photos and art.
Describe works like any other Midjourney command, but instead of /imagine, a creator uses /describe. After entering the command, a drop/drag box will come up, letting the creator upload a picture they want to use as the basis for their creation.
For this example, we used a famous picture of John Whiteside Parsons, co-founder of the Jet Propulsion Laboratory at the California Institute of Technology (CalTech). Parsons was an enigmatic figure who was also a follower of British occultist Aleister Crowley.
After the image was uploaded to Midjourney, the AI gave four options to generate the next picture. For this example, we chose "an old photograph of a man holding a wire, in the style of gritty Hollywood glamour, synthetist innovator --ar 16:9."
Midjourney then produced four images based on the prompt.
From there, we could make variations, upscale, remaster or download the image.
An interesting footnote: Midjourney seemed to cure a long-standing problem it has: it can’t convincingly draw hands. Despite incredible leaps in artificial intelligence, generative AI platforms continue to struggle with creating human hands. Experts attribute this to a lack of data on the human body that includes images of hands given to the program to process.
And voila! The AI appears to be learning. Very Cool. Or… creepy?