At the heart of new artistic communities, images created by artificial intelligence fascinate and worry

All day, the servers of the Discord community dedicated to the Midjourney image generator tool see thousands of request messages pouring in. On this server, which has more than two and a half million members, all you have to do is type in a text description for a bot to return a series of four images. There are abstract paintings meant to depict the last human being on Earth, very specific queries (like ” cats shooting lasers in space, realistic, 4K, cinematic ”) or on the contrary very vague (like “blockchain” Where “invent reality”). Whatever the request sent by the user, as long as it does not contravene the regulations which in particular prohibit the production of content ” visually shocking », images will be generated with more or less satisfactory results.

Since June, software capable of producing images from a line of text – based on a database that can include hundreds of millions of images from the Internet – has exploded in popularity on social networks, largely thanks to the sometimes poetic, sometimes surprising visuals that they generate. While it is fairly easy to produce an image using programs like Midjourney, it takes some time to figure out what kind of turn of phrase to use to get a good rendering, or at least faithful to what the user expects. – a process called the prompt crafting (“instruction making”).

To get this image in Midjourney, you had to type the query:

The art of query

In order to improve the images they produce, users therefore frequently exchange advice in order to refine their textual descriptions or circumvent the weak points of the artificial intelligences used. ” Looks like there’s a bug with redhead eye color. We’re experimenting, but it’s impossible to give them anything other than blue or green eyes at this time. laments one Midjourney user, while another begs other members to help her find the right keywords to generate an anatomically correct lion. On a Reddit thread dedicated to the Stable Diffusion toolthe same questions are raised: “ Who knows how to produce coherent human beings instead of these abstract monstrosities ? ” Where ” Anyone have a good description for spawning a goblin? “, to recite nobody else but them.

Some users have put together free guides to help community members get to grips with these different programs. In a digital book dealing with DALL-E 2, another software, we find in particular adjectives easily understood by the tool, as well as examples of descriptions to generate realistic images, paintings or even scientific diagrams. ” Every day, I saw interesting discoveries on Twitter, Reddit and Discord, and I wanted to collect them in one place “explains Guy Parsons, author of the book.

You have 59.68% of this article left to read. The following is for subscribers only.

We want to say thanks to the author of this article for this outstanding web content

At the heart of new artistic communities, images created by artificial intelligence fascinate and worry

Our social media profiles here as well as other pages on related topics here.