From Midjourney to Crayion, every AI system we tried has a number of free trials. Each program gave its own personal interpretation to the text.
Producing works with artificial intelligence is a bit like cooking carbonara. Everyone knows how to do it, some are very successful. The first steps are intuitive even for beginners, you enter the sites, read, orient yourself and experiment. All of them have a limited number of free trials (5 to 10). So, rest assured, those who manage to produce aesthetically passable images on the last attempt are already good.
The software is similar but they are not all identical. For example, by inserting as text: “Dancing meerkats toasting with Bacchus during a storm”, each program gave its own interpretation to the text. After a few tries we want to give some advice. The more information you enter, the closer the job is to the desired result. It is not necessary to put only “narrative” information but also stylistic information, for example by inserting a reference to a painter or an artist with a recognizable aesthetic.
Combining unusual terms or using evocative phrases can instead be a ploy to avoid ending up in front of the first Google image result. “Girl in a meadow with a flower in her hand” is wasted proof of this. To avoid. The order of words in the text can also affect the dominance of one image or concept over another. Sometimes you can just try to rephrase the sentence. However, these are just basic tips. Working with artificial intelligence is more complex. To do it well, you need to subscribe to the best performing software.
All the software to start creating images with artificial intelligence
To use Midjourney, you must create a Discord account. Here, in fact, text messages are sent to the bot which connects to the artificial intelligence. Once entered, you must join the channel and “accept the invitation”. There are several rooms, to take advantage of the free trials you have to enter the test area. Once inside, just enter the /image command and start the image description. Strictly in English. If inserted in Italian, it often manages to capture just one of the words.
At this point, Midjourney presents four possibilities and the author chooses one of the images. With the U button can create a larger version of the work, with V, instead, it generates a new image based on the selected one. If the author doesn’t want to choose any of the four images, they can always click the dice icon which generates four more images from the same prompt (the entered command). At any time, however, you can save images by selecting them or by tapping the download icon on your smartphone.
Also on Discord is Bored.Ai, which resembles Midjourney. Also in this case, just enter the #generator section of the channel. To get started, you need to enter the command !Go, followed by the text. Bored.Ai, does not immediately offer the images generated by the AI, but as a first answer it provides four texts. Four more detailed interpretations of the description entered by the author. For example, he asks whether or not a figure should occupy the center of the painting, what is in the foreground and what is in the background.
Generally the answers are quite adherent to the composition of the image, sometimes, if the sentences are more cryptic, the software can paraphrase the text in a different way and make it take on new meanings. Once one of the interpretations is chosen, Bored.Ai generates four images. You can select one or, again with the relaunch key, choose to start over.
Beyond these options, there is another simpler but less powerful one. With ArtBreeder, just enter the site and click try demo, write keywords and send the information to the software. Unlike Midjourney or Bord.Ai, ArtBreeder does not require an intermediate step to produce the work, nor the possibility of modification from a first generated draft. There is only the reroll button which makes it possible to process a different image from the same keywords.
Night Café Creator is a bit like ArtBreeder, you enter, write the keywords and wait for the AI to generate the image from the text. In reality, however, there is an intermediate stage, to frame the final result from a stylistic point of view. In fact, before sending the order, it is necessary to select a filter (oil painting, sinister, horror, comic, etc.).
Starryai, on the other hand, allows you to choose between two creation modes: a more abstract one, Altair, which uses a VQGAN-CLIP model, and a more realistic one, called Orion, which uses a CLIP-Guided Diffusion model. Then you get to the panel where you can type the input text and choose the reading style from a predefined list. Crayion, also called Dall E mini, through a spartan interface allows you to enter the desired text. It works well with simple words. Speaking of “dancing meerkats toasting with Bacchus during a storm,” Crayion interpreted it that way.
We wish to thank the author of this article for this amazing content
We Created These Ten Paintings Using Only Artificial Intelligence
Discover our social media accounts as well as the other related pageshttps://www.ai-magazine.com/related-pages/