Two serious warnings to start with:
NSFW (Not Safe For Work), i.e., may contain sexual, violent or another adult motives, at least that's what any hypersensitive people say
Generating images with AI is an engaging occupation, a time thief. Look out!
The planet is changing faster. I inactive remember erstwhile the only examples of artificial intelligence were dyed blondes. And here's a large change, utilizing the AI available to the people. You can't get over it. In the authoritative post office there are incentives to usage AI in writing works. To be honest, most of the current work already looks written by AI, so I'd alternatively usage the reading device. large advocate of the 4th industrial revolution, transhumanism and digital enslavement of humanity, Oberführer WEF Klaus Schwab in his book 4th Revolution Industrial claims that most of the texts are written by AI. This may not seem credible, but let us remember the thousands of pages of various bullshit manuals, messages, letters to customers, etc., all this gibberish, which is now the majority of the written word. Also, looking at the productions of journalists without any trace of criticism and self-thinking crying choir under the dictates of the directors, 1 can be convinced that even if they are inactive surviving people, they can be replaced without any harm to the quality of journalism or even to the benefit of it. fresh confusion around the AI (reported as Sam Altman) came from the fact that any of the project's leaders did not like the freedom of expression of the AI. The media did not show it (a surprise), but the actual controversy was whether and how the AI should be censored. For our sake, of course. Freedom of speech is simply a threat.
However, it was expected to be about generating images. This is served by various alleged deep learning models that can make images based on text descriptions. A peculiar model is Stable Diffusion, due to the fact that it is mostly available and its origin code is open. This besides means that you can usage it free of charge. It is said that with a sufficiently powerful GPU (Graphics Processing Unit) home computer, you can besides work without net access after installing the appropriate programs. There are a large number of pages on the network that both describe the model itself and work with it. There are besides descriptions of the installation process, but in order to make images we do not gotta install everything at home.
There are presently mostly available portals online to make or process images utilizing AI. utilizing them has the advantage that we do not gotta install anything, and we besides have the anticipation of utilizing other, more advanced models with more possibilities. You can usage different types of specified services, especially the generation of images based on text. Access via web browser. Who has good eyes and skillful fingers can do so even through the phone. Here I will describe my experience with the usage of NightCafe for respective days. It is 1 of the popular access portals. The environment to work through the website is Night CafeStudio (NCS) https://creator.nightcafe.studio/
Using the NCS requires free registration, as usual for this intent it is essential to enter an email address, choice a password and you can already have fun. The creation of images is mostly not free, but points, called credits, are awarded for starters. On a good morning we get any of them, then after 5 each next day, just log in all day and click Claim your regular free creditsTo keep them from disappearing. erstwhile you request much more, you can buy more, for example by choosing any form of subscription, but mostly it is not necessary, due to the fact that we get further credits for different activities, specified as publishing our works, judging or liking others, etc. We besides appreciate our work, that is, the likes and comments of another authors translate into additional credits.
Most importantly, we can usage a free model and work with small resolution. Then the generation of the image costs nothing, and only if the consequence is considered worth developing or wanting to usage somewhere, can we improve its resolution as loans. The models utilized are free of charge in the case of NCS already mentioned Stable Diffusion and Dream Shaper. Since the credits are not lost, we can usage the saved credits to later usage another models or to process work with more resolution, and so far we have fun without paying.
All images shown here were generated without paying utilizing the model DreamShaper v8 and at minimum resolution 512*512 px (Thumb) and were not further processed in any way.
The generation of images is simple. utilizing the mouse, we choose a model (preferably a free one) and a predefined kind (there are many, but you can besides not choose any). In a window entitled Text Prompt We enter a description of the image we want to get, at least 1 noun. And in this, that is, almost the full art is included in the shaping of the promptium. I tried in Polish, but everything indicates that AI does not talk this language, so you request English. I've besides seen French prompties. This is not a large problem, due to the fact that simple statements needed for this intent can easy be obtained with a Google translator. In Bookmark Explore We'll see what others have generated. Many authors of the work hide their props, but any can be read to familiarize themselves with examples. To begin with, however, we will go into deep water to make respective images utilizing a free model. Note that by repeating the generation with the same prompt, we get different results. There is nothing unusual about it, for our description is not exhaustive and leaves a large deal of freedom. It is interesting to compare the results obtained for different styles. Moreover, the nature of the images depends powerfully on the chosen model. For example, the SDXL 1.0 model, the most popular NCS model, imprints its clear mark and so I do not like it.
I chose the Dream Shaper model, the no style, that is, the deficiency of a predefined style, and I entered a profoundly thought-out and highly well-developed prompt consisting of 1 word “fairy”. I ran the show twice, generating 2 pictures of gorgeous beauty. Here they are:
Truly, I tell you, erstwhile I look at the works published by others, I fishy that most authors hide their prompt not due to the protection of their large contribution of work and sophistication, but that it may not seem so trivial.
Our small circumstantial prompt will make all kinds of fairies, so by moving the process many times, we can look forward to the desired result, or many of them. We will act more ambitiously and complicate the prompt by introducing an additional term. We add 1 word to the promptu “breasty”, we now have “breasty fairy” and a fresh gallery of images. It would not be possible to compose “fairy with boots” due to the fact that specified a prompt is blocked as containing a forbidden word. We tricked the system, but nothing unsuspecting audience could be electrocuted. Fortunately, the fresh images will be marked by the strategy as NSFW (Not Safe For Work), as the song reportedly contains sexual, violent or another themes for adults, from which society should be protected. Horror and terror.
Especially not amazed that erstwhile we look at another people's works, we will find quite a few fantasy characters from movies, books and games. The most are warriors, fairies, wizards, sweet kittens, pink ponies, mysterious landscapes, etc. However, if we have more crystallized requirements, we request to work on the prompt, even if only to match the consequence to the character or scenery of the favourite game.
I mentioned pre-defined styles. There are quite a few them on the NCS, for example 2 of them, cubism and surrealism. There are only 2 pictures in this gallery, but I had to make many of them in order to choose something interesting. This is getting interesting. Here are the selected (of course besides NSFW):
Pre-defined styles are over 40, but they don't gotta suit us. Fortunately, you can specify different styles, for example by referring to the painter’s name. The fresh prompt will look so “breasty fairy, XYZ styles”, where XYZ is the painter's name (or name) or artistic kind name, or era name. Of course, it is good erstwhile the painter's works are in line with the subject of the image created. The kind of architecture artist or conflict scenes is poorly suited to female characters, AI will not find much useful information. The image will be generated, it may look good, but it will not necessarily be in the right style, AI can do something like a fake kind or skip it completely. Of course, this is not always a problem.
If for any reason our entry is not identified, And I won't interpret him, I will quietly ignore him. For example, due to a typo or erstwhile the artist is missing from the database. Sometimes it just happens for no reason. The command is ignored without comment due to the fact that AI is not just a machine, it is simply a personality. The ability to ignore commands is an intelligence at least at dog level. I've hosted individual like this for a twelve years, so let me tell you something.
Unfortunately, there is simply a shortage of Polish artists from fresh times like Daniel Mróz or Tadeusz Kulisiewicz. artist Zdzisław Beksinski is, but possibly due to the fact that he was well known in the West? With another contemporary painters I have not yet tried. But I know that they don't should be painters, due to the fact that I found in another authors the illustrators of books and architects.
Below are any examples. Each time it is simply a choice from respective generated images.
Picasso
Rubens
Malczewski
Fly and Klimt
Edouard Manet
I made more pictures utilizing fin de siècle styles, Witkacy style, sumi-e style, Utagawa Hiroshige style, 1920s russian Union kind (you gotta give a decade due to the fact that the kind has changed), 3rd Reich style, realistic photograph (with different attributes) and a fewer others, but I will not present them here anymore due to the fact that it is getting boring. You can proceed to have fun, but I think specified a example is adequate to give an thought of the possibilities of generating images based on text. It is possible to specify your own styles, but it is already a higher school of driving. I didn't show it, but shaping the promptue allows for a alternatively detailed definition of the image. In addition, the prompt may have respective parts, which are given different weights. The combination of promptu and kind selection gives a powerful creative tool. There may be imperfections, specified as the unusual form of the hand, or the fingers bent at a unusual angle, but who would look at the fingers of a deep fairy? There's a 3rd leg, too. This can be solved by shaping the prompt.
It would so appear that to successfully make images utilizing AI no longer needs to have any thought of plastic arts. To any extent, yes, but it's not precisely true. We kind in the text and we enjoy it erstwhile something good comes out. We're 1 step distant from complacency. However, a certain concept and sense are needed for a meaningful selection of results. From my small week-long experience, but besides observations of respective interesting authors at NightCafe, we print at most about 10% of our own productions. Also, to compose a sensible prompt, it's good to know what we want to get, and that's what we request to learn. You can experiment, but it is besides worth looking at the achievements of others. And a choice of style. You request cognition of art to find it or to make it suitable for your needs. You can combine elements of different styles, but first you gotta know a lot about them, you can identify the key qualities of the kind that we want to introduce to the description, etc.
In my opinion, the current revolution in art is somewhat like what happened in radio and electronics. There utilized to be systems made of individual discrete elements, specified as tubes or transistors, resistors, capacitors, etc. They were designed and built again each time. Then there were integrated circuits, and erstwhile an amplifier or something is needed, a corresponding integrated amplifier or another scalak is selected from the catalogue. The layouts become more complex, and the creator does not deal with details, but with the full work. Undoubtedly, the device will master a large part of its use. Already now, most of the commercials and movie setting are created in this way. Do you think these happy families and talking laundry pills from tv commercials are real? I think that the lamentation of the fall of art is premature. Let's look at the most productive artists in the past. They utilized many devices to facilitate and velocity up work. Leonardo invented the painting roller. Rubens and many others had a squad of apprentices. I've seen sculptors usage mechanized tools a long time ago, and lately I've seen them usage laser cutting. There's another powerful tool here. However, there will always be area for the actions of the artist's man who will have more tools at his disposal.
I'm planning another part where I'll show you any of my work.
Beksiński
Thank you for reading Substack Jack! Sign up for free to receive fresh posts and support my work.
