@Ciaran
It works for normal logged in users (other arguments do not and are staff only).

@The Smiling Pony
Ah! Perfect - That makes sense. Thank you.

Yeah I had a few rule 1 violation takedowns myself, mostly because i used “rwby_team” in the prompt which slightly altered it, not much but it did. The reason was they didnt want any issue by roosterteeth. A bit weird considering at the moment its a bankrupt and dead company unless they find a buyer but I get the intent. Only reason I used that term was civitai doesnt like equestria girls heads, it keeps thinking they are underage, regardless if they are 8 feet tall with huge breasts, its an automatic thing to do with eye/head scaling. Used to happen to my princess peach ones but after a few thousand images I think they know they could ignore those or they tweaked their settings where it doesnt flag her anymore idk.

I previously made some big mistakes uploading images here, so I may ask questions that have obvious answers. For a quick sanity check, for rule one would you say that an “renown dead painter” is of “renown” if they have a Wikipedia entry?

@AiBoi
We look at images case by case. Most of that stuff in Rule #1 is to protect human fan artists, not commercial properties which have their own set of “don’t do this” which, without a legal team, is something we handle when it happens.

If you’re going to recreate something by Michelangelo but with ponies, that’s the kind of thing we already have and it’s ok. It’s not a problem.

Beyond that it’s hard to speculate, other than to say that if they’re famous enough that they have a Wiki article, then they would fit the “highly commercial or culturally significant styles (e.g. from renown dead painters)” clause of Rule #1. But, at the same time, if they’re a recently passed Disney artist or the image recreates something from a famous Hasbro artist, wiki article or not if the image you’re recreating is still something that some company like LionsGate has a hardon for, that might be it’s own kind of problem.

So, without seeing what you’re doing it’s hard to speculate beyond “Sure, probably. Let’s see.”

@Ciaran
Thanks for the detailed response! I completely understand why it has to be viewed case-by-case view, I just want to develop a couple of simple “rules-of-thumb” for guidance. I understand that it could be wrong at times, but I’m hoping for getting it more of a “We think you have strayed outside the lines” rather than a “The line is not the start of a long jump sand pit” if I get it wrong.

I’ve got a couple more questions.

1: I have serval images I’ve posted to other sites that don’t follow rule #1, but I’m thinking about reworking them so they do. Is a simple img2img pass into an art style that is in compliance good enough, or is that just “laundering” the dirty image like dirty money? If so, would a txt2img generation using a controlnet be ok, or would it run into the same issue?

2: For “highly commercial” a quick test I thought of is if you mostly purchase/consume the artists work under their name vs. a products name. For example, Rumiko Takahashi is the lead art designer for several very popular animes and mangas, and you consume and enjoy her works via that commercial medium, rather than directly, and would likely be ok to use.

Compare that to Ayako Rokkaku, whos art you do not consume via another name, such as a TV show, game, comic, etc., but rather more directly.

In short, when you see Takashi’s art, you tend to say “That’s a image from Inuyasha” and when you see Rokkaku’s you say “That’s a Rokkaku piece.” Does this sound sane?

3: If there’s some indication that an artist is ok with the use of AI image generation, what would be the best way to convey that information to the mods? For example, I found this article that, amongst other people, has some quotes from Mike Winkelmann, a.k.a “Beeple.” While the cover image is a big giant anti-AI image generation image done by him, at the end of the article there’s a line stating “His advice to up-and-coming digital artists is to do what he did: use the system as much as possible, and lean in.” To me that sounds like he’s ok with people using his stuff for generation, but I don’t want to get it wrong here.

Once again thanks for your time, and sorry for the wall of text. I was very wrong once, and I don’t want to be again.

If you take an image that was specifically intended to emulate an artist’s style, as almost all your uploads have, and you pass it through img2img, then you’re still emulating that artist’s style but with extra steps.

If an artist gave you permission to emulate their style, tag them since your work will be derivative and mention that you have their permission in the description.

@Ciaran
Got it. I’ll step all the way back to txt2img then. I’ll also assume that controlnet guidance would also be in violation.

If you’re using it for stuff like HED or SoftEdge or depth or color or map, of course not. If you’re using it to emulate a specific artist’s style, then given your other uploads almost certainly.

Please just stop emulating fan artist’s styles.

@Ciaran
Would you believe that AI generation actually lead me into getting commissions, and lead to a bit of spending spree? Before I got into it, the ideas that I could get an image to my specifications felt foreign and the prices felt too high. After getting images more exactly to my liking I realized that it felt better than just finding something like what I want, and I became more open to the idea. I’ve also spent hours doing less work than the average artist to make an image, which really helped put the amount of work they put into their works into perspective. I even used this site to find about half of them.

The other sad thing is that the one artist I most want to commission, yakovlev-vad isn’t open for commissions. I know that fact doesn’t give me permission to emulate their style, this is just me moaning.

Uhhh…

“Lead art designer” Rumiko Takahashi has a pretty distinctive style in any case.
She’s arguably the biggest manga author/artist in Japan, but whatever.

Wouldn’t it depend on the denoising strength, though? With this value high enough, the end result might be very different from the input image.

@tyto4tme4l
If the prompt changes to avoid that style and the denoising is high enough to effectively “erase” it and just keep the image concept (pose, items, bg, etc.) then that should be fine.

The intent of the rule is to prevent people from taking away what makes a certain artist “unique”, the style they worked to build up and use as their identity.

I feel like I’m getting conflicting information on this stuff. For example, I was told directly that use of artist tags counts as emulating an artists style, even if the resulting output doesn’t really mimic the artists style. I also have been told that the use of “…existing styles to composite something unique that doesn’t look like anyone’s specific style should be fine…” Then we just had what frowning just said compared to Ciaran.

@AiBoi
When your prompt specifically lists a fan artist whose style you want the ai to emulate, that’s going to be Rule #1.

And when it’s clear whose style you were trying to emulate just looking at the image, it’s also Rule #1 - even if you think you “didn’t really” mimic the artist’s style or somehow should have “gotten away with it” because you specified two or three other artists in the same prompt.

Many of your uploads haven’t been deleted, in many cases because of exactly what TSP said. But many of them were deleted because they were absolutely and very intentionally emulating specific artist’s styles.

2LDR - on your prompts that you uploaded that I was involved with, you weren’t just “keep(ing) the image concept (pose, items, bg, etc.)”. People were literally commenting on some of your uploads how well you emulated the artist’s unique style.

You’ve started using dead and non-fan artists in your prompts. Those aren’t a problem. Why aren’t you ok just continuing to do that?

I don’t mind that I can’t use fan artists, frankly branching out into other “mediums” has been fun. It’s just frustrating to have one person say X and another Y. But, as I said, I’m going to stay away from fan artists. It’s also frustrating to look at an image and know that artist tags had to have been used, but not know which ones. I would report them, but what am I going to say? “I feel like this had to be done using E621 artist tags due to the model, but I don’t know which ones” don’t feel like an effective report. I’ve checked metadata on a couple, and the tags are clearly there. Should I just report it with “Metadata contains the following artist tags: x, y, z”?

Also, I feel like the worst offending images came from third-person upload. I know of a least one image that was a dead ringer for a fan artist that was uploaded by a third party. I should have been a better person and reported it myself.

your quite rude.

ehh, that seems like its placing a double standard on AI, a double standard. Humans can make pastiches and homages to other artists, and outright copy their styles for inspiration, but AI can’t.

@Background Pony #FBCF
Since when is copying a style part of being inspired? Sure, you could see a minimalist drawing and think “I could try something like that”, but there’s a difference between that and actively attempting to copy that artist’s specific style. One is inspiration, the other is dishonest.

Also, please, the generators are not in any way comparable to humans. They’re just mixing training data together based on the tags you input.

@Parallel Black
I’d like to know what you think humans are doing when they make art if not “mixing things you’ve seen before based on desired attributes”. Unless you’re claiming all modern art just comes out of a vacuum and could be created without looking at a single piece of existing art or something.

Can we please not debate this again.

How is the garbage pile doing these days? Still increasing steadily or has the rate of accumulation gone up?

@Background Pony #CD82
Considering the imposed upload limitations, it hasn’t changed much.

@ZippySqrl
It’d be nice if the APIs here made it not so troublesome to break down that graph by rating. My anecdotal evidence is that the percentage of new clop images that are AI is much higher than the amount of new SFW AI art.

@ZippySqrl
@🐴
You might be interested about this feature request post I just made.