Portraits Replicating Styles of Photographers Made with DALL-E 2

Take a look at Photographer Michael Green's experiments with the text-to-image tool DALL-E 2. Green tried to replicate the styles of well-known portrait photographers with the help of AI. 

In each instance, Green used the phrase "Photo of a young Mexican woman in the style of [photographer]." For each artist, he completed 2 or 3 iterations, with each iteration returning six unique images, and then chose the best 4.

The photographer noted that DALL-E 2 has "an incredibly refined understanding of lighting and composition." When the source artist has a consistent style, the AI was able to replicate it with accuracy in some cases.

Green also noticed there might be content filters after the images are generated that scan for nudity.

The artist thinks DALL-E 2 has the most difficulty with eyes as there are a lot of glitches, but then sometimes it renders with remarkable detail.

Green also tried to replicate Vivian Maier's style, but the results came up "with a freaky camera."

DALL-E works really well for some photographers but struggles with others. For example, the Gregory Crewdson style shots are thematically correct but the details of figures are "glitchy."

Check out all the works in this thread and read this article if you want to learn more about generative art.