Tool recommendation to convert pictures into cartoons

[ChaosPriest]

I want to try making art using actual photos and running them through some kind of filter to turn them into a cartoon. It seems like there are a lot of options out there, but I don' t know how to sift through the SEO garbage.

Can anyone recommend a particular tool for this job? Is it reasonable to expect some kind of open source version to exist?

 

[Winterfire]

Outside of AI (Stable Diffusion with cartoony models, img2img), normal filters are rarely if ever good enough for an average result, especially if you do not want to put any manual work into it.

 

[anne O'nymous]

I want to try making art using actual photos and running them through some kind of filter to turn them into a cartoon.

This being a piracy site, the point is relative, but hmm... Do you own the rights for the said "actual photos" ?
While turning them into cartoon would probably give you some right over the creation, using photos/visuals you own no rights on can lead to a lot of trouble ; from a strong incentive to stop immediately to having to face a judge.
Games like, by example, Good Girl Gone Bad, are safe, because the author trace over the pictures. So, it's a full creation "inspired by". But what you want is to use the photos then apply some filter(s) on top of them, what is radically different.

If it's generic photos that you want to use as background, there's few risk. But if you intend to use more than background, counting on the "cartoon" effect to not looks like you used real images, well, do it at your own risk.

 

[ChaosPriest]

I agree that such a transformation is not a way to evade copyright. In fact, even tracing over pictures might still count as a derivative work. I do like the idea of tracing over pictures.

I am more interested in changing the style of the art, and maybe being able to use photos of two different models that appear to be the same person once they are converted.

 

[drapak12]

There were some list compraring cons/pros of most popular ai generators. I do not remember all which I tried but I found pixai.art.
Many option to adjust cartoon style and transformation level. Free if you are not hurry. Sometimes generates hands with 5 fingers but does not like genitalia.

 

[osanaiko]

I want to try making art using actual photos and running them through some kind of filter to turn them into a cartoon. It seems like there are a lot of options out there, but I don' t know how to sift through the SEO garbage.

Can anyone recommend a particular tool for this job? Is it reasonable to expect some kind of open source version to exist?

Here's an example with some art transformation I was tossing around to use in a project idea.

Left is a daz render, right is one of the better results after spending about 30 mins messing around with Automaticc1111 img2img. Note that I've already spent a fair bit of time learning how to use the stable diffusion style tools to a basic level, so the learning time is not *that* fast, but it's still years faster than learning to draw

Yes, I realize there are imperfections with the eyes and the hands, but it's close enough even my limited photo edit skills could fix it together with more StabDiff infill generation.

One thing I *DO* like about using a cartoon-style trained checkpoint is that you don't get the distinctive "AI hyper realism", which I find very distracting in the games/images that use more blatantly generate images.

 

[Count Morado]

Games like, by example, Good Girl Gone Bad, are safe, because the author trace over the pictures.

I would disagree. Although this is a question for the copyright courts in the respective countries in which a suit could be filed, in the US SCOTUS has already ruled on something similar (though not on all-fours) through their decision in Andy Warhol Foundation for Visual Arts, Inc. v. Goldsmith, 598 U.S. ___ (2023). GGGB may make enough changes in some images, but if the similarity of the humans traced is too recognizable to the original imagery, a court of law could consider it not enough alteration for "fair use."

Here's an example with some art transformation I was tossing around to use in a project idea.

Left is a daz render, right is one of the better results after spending about 30 mins messing around with Automaticc1111 img2img. Note that I've already spent a fair bit of time learning how to use the stable diffusion style tools to a basic level, so the learning time is not *that* fast, but it's still years faster than learning to draw

Yes, I realize there are imperfections with the eyes and the hands, but it's close enough even my limited photo edit skills could fix it together with more StabDiff infill generation.

One thing I *DO* like about using a cartoon-style trained checkpoint is that you don't get the distinctive "AI hyper realism", which I find very distracting in the games/images that use more blatantly generate images.

View attachment 3402903

This could have a higher chance to pass muster for "fair use" transformation even though it is obvious that the generated model is based upon the original, there are enough differences that it could be considered a new work. However, since it would be created mainly by AI generation, there may not be enough human editing involved for it to gain copyright on its own.

Again, your country's laws may vary and if a person is worried, they should consult an attorney who specializes in copyright law with visual transformative works.

 

[osanaiko]

if a person is worried, they should consult an attorney who specializes in copyright law with visual transformative works.

Or they could just YOLO it because this is just a porno games site and the chances of making real money and coming to the attention of the rightsholder is very very close to zero.

 

[Count Morado]

Or they could just YOLO it because this is just a porno games site and the chances of making real money and coming to the attention of the rightsholder is very very close to zero.

It's farther from zero than you think. But yes.

 

[anne O'nymous]

I would disagree.

Yeah, giving it a bit more thought, I should probably have said that it's "more safe" for games that are using tracing. Everything depending on how near to the original the result is, and how much you stick to a single model.
If you trace because you're bad at posing, using the source because of the global shape and pose, and then trace again, from another source, to the face expression because you're also bad at this, the risks are lower than if you trace the whole thing then color it with the same scheme, using systematically the same model.