I was going to post this question in a new thread but after doing a little search this seems like the right place to post it, even if a lot has changed in 2 years.
Is the community opposed to someone posting a plugin, or attempting to contribute a feature which primarily came from asking one of these ai systems “write me a krita plugin, it needs x, y, z features, etc, etc”.
In the coming years these ai systems may allow an entirely non-technical person to solve problems the current krita contributors face within mere hours or days, rather than weeks or months.
It’s plugins that use AI methods to generate artwork that the forum is opposed to.
There may be further issues where AI assisted methods of writing code/scripts have problems because of copying the work of others without giving credit for it.
I know that human coders do that all the time
I think that would be a separate discussion from the artwork discussions.
You can create whatever plugin you want with an AI, if an AI can do it.
Just take care about license compatibility of produced code.
If you’re creating code with an AI that has been trained with copyrighted source code or open source code like GPL, then you have to care: the code produced for your plugin may violate the license from source code from which the AI has been trained on…
It’s exactly like most of artworks, musics, articles or whatever stuff generative AI produce: it’s built using copyrighted data without any consent or retribution of their author
Hey, just wanted to link this post by @Alien_Sunset as the original thread is already very busy. It shares a link to a new AI model to create fake timelapses.
Wow, I was shocked by this… This is disgusting and got a fair share of backlash on GitHub, but probably it won’t get taken down anyway. This is pure malice, a tool to make scamming easier than before.
Ugh. I know the online comments are just a minority, but it really gets to me how these people seem to only have contempt for artists.
I guess this new model cannot fake the whole painting app UI (yet), but if a person doesn’t know how digital painting works, I doubt they will be able to tell the difference anyway
Thanks, and yeah, I was afraid something like this may be the case. That if authors of Glaze would just publish the source, it would be trivial to bypass their protection when training the models.
In other words, the search for a robust defense mechanism continues.
I’ve recently stumbled upon the very sad news. AI now can generate art creation time-lapses, and even though it’s a bit sloppy at it right now, it will definitely improve in the nearest future. At first, I thought it was just some really enthusiastic AI bros doing this, but it appears that time-lapse-capable ML is being developed by a real research institute. To me this looks like war. War on real human artists. It is highly unethical, but I guess our destiny was already decided by the kings of the Silicon Valley. Unless we do something to turn the tide! https://www.csail.mit.edu/news/using-ai-recreate-how-artists-painted-their-masterpieces
I don’t quite understand the purpose of this feature. I can do sketches myself and people who don’t draw will be interested in the finished product, not sketches.
As I understand, Paints-Undo can create a fake “time-lapse” video for AI-generated images. This way, people who merely write prompts will be able to convince others that they have drawn/painted this generated artwork themselves, even though it’s not true. In perspective, this will create a nightmare situation for real artists because showing your art process will no longer be a proof good enough. Some real artists are attacked already, because their art is “too good to be drawn by a human”, and those artists still could prove their authorship by sharing a time-lapse that showcases their art process. But this will change soon… Which means we need to create some sort of real artist certification system that can’t be faked or taken over by tech- and political-activists.
Yep, now we need a video, filmed from behind over the artist’s shoulder, showing the artist’s painting their pictures with their own hands.
But how long will it take till supercomputers are powerful enough to fake even the whole process filmed in the way I suggested?
In fact, people tend to become suspicious of everything. The other day I was watching a video of this little girl singer.… and some of the comments said she was “too perfect” and maybe it was the work of AI.
I was interested to know how you think Ai can help generate color palettes for digital artists? Say getting from a text input command from the artists (can be any high level command, even a theme), and have a color palette or color scheme generated based on the prompt. What do you think would be cases that can help the most and facilitate color selection for artist. Im thinking of ways that AI can assist …
How would you describe color to an AI (or even a person) in a meaningful way? What could AI offer when you’re looking for a theme that a simple search engine can’t? Colors are a lot about feeling and you try to outsource your feeling for colors to a machine that has no understanding of these concepts.
Exactly… Thats what I like to know. We all have different perception of color, and I wanted to see if we can use high-level descriptive language to tell AI to create a certain composition or color scheme, certain mood, etc… Do you think it’s a possible thing that artists would wanna use? Also things like blending a color on different layers, or fixing its brightness, saturation etc based on a user command or other elements and color compositions on canvas …
This sounds like way to imprecise. It will always be easier and faster to just pick colors from the color picker or a reference image and for blending just use a blending or smudge brush. It doesn’t make sense to write a paragraph of text for things you can do in a few brush strokes or clicks. It’s a waste, especially when you think about how much computational resources LLM AI needs for even simple things like this. Sure you can tell generative AI to create an image in a specific mood or with specific composition but then you have just the same that is was current AI already does, it generates a finished image. Nothing really useful when you are still actually a real artist who wants to paint themselves instead of retouching instant images.
