For some time now, many people have been experimenting with the amazing possibilities of artificial intelligence in text-to-image. Don’t you know it? Well, yes, the press articles these days are all about DALL-E and the funny images its journalists generated by simply writing the illustration they were looking for and, eventually, the style.
Would you like to try? Go to Craiyon.
It is quite tempting, rather than going out to find the right image to document your article, to ask the AI to make the ideal illustration. Let’s see how Craiyon imagines it…
What about DALL-E ?
The principle? For years, the AI has been trained to recognize objects in images and extract words to describe them — the “image-to-text”. Here is the reverse operation: we give words, the AI synthesizes images — the “text-to-image”.
Recently, a photographer, Michael Green, tried to ask DALL-E to create images in the style of some great photographers and he posted the result on his Twitter account (Triplux).
The phrase he typed was : “Photo of a young Mexican woman in the style of” (Dorothea Lange, Helmut Newton, Diane Arbus, Eric Lafforgue, Annie Leibovitz, Lee Jeffries, Steve McCurry, Nancy Goldin, Robert Mapplethorpe, David Bailey, Terry Richardson, Yousuf Karsh, Martin Schoeller, Julia Margaret Cameron, George Hurrell, Peter Lindberg, Brian Ingram, Kenneth Willardt, Richard Avedon, Dmitry Ageev, Rosie Matheson, Herb Ritts, Arnold Newman, Robert Frank, Bruce Gilden, Vivian Maier, Gregory Crewdson).
I don’t repost here the images he obtained (sometimes at the cost of several attempts) and choosed, and that he comments briefly in his tweets. Even if the AI is probably the “author” of the pictures (what does this word mean, then?), Michael Green is the curator and, in a way, the co-author.
I’m just using the idea with other photographers (without doing the necessary curation work, and you will see that it is necessary — the images obtained are often disappointing) to illustrate the point. And I await your own comments…
