Hayao Miyasaki is the co-founder of Studio Ghibli, a Japanese animation studio known worldwide for their stunning, emotional, beautiful stories and movies. At the core of Studio Ghibli’s work is a deep engagement with questions of humanity. About what it means to be a human, about how to care for one another and the world […]
Did they specifically allow “Ghibly style?” Or did they just loosen the restrictions on asking for styles in general, and Ghibly style just turned out to be the popular one that memes started snowballing around?
For the longest time OpenAI’s systems would try to block people from generating images in the style of certain artists. This was obviously for copyright reasons, the didn’t want to get sued (even more than they already are). Which is something they just changed very explicitly. You can now easily generate stuff in the style of Studio Ghibli and Sam Altman made his avatar on X-The Nazi Network a ghiblified version of himself.
I don’t have specifics if they have allowed other styles to be used now, too. I don’t use this nonsense, but it’s clear that Ghibli was put front and center.
Yes, I read the article. But it doesn’t answer my question. Did OpenAI specifically enable Ghibli style, or did it remove the restrictions in general?
Everyone’s pulling out Miyazaki’s out-of-context quote about procedural animation and are interpreting this as some kind of personal attack against him in particular because of it, but unless OpenAI specifically made Ghibli style available without lifting restrictions on others I don’t see a reason to assume that.
Also, an article that calls X “The Nazi Network” is not exactly the most reliable source. This isn’t even about X.
That didn’t exactly look like animation. Looks like they trained an AI to control a humanoid figure in a virtual environment. It learned completely new and inhuman means of locomotion. Not very impressive from the technical angle, but the pitch about using it as a model for Zombie movement was clever.
You can use that for CG animation, of course. But those bi- and quadrupedal robots are also trained that way.
I feel the filmmakers manufactured some drama there. Knowing the real context of the quote makes it much more sensible.
This suggests that all they’ve ever actually been doing is blocking keywords of artists names, and that it has always been trivial to get around such restrictions if you know how to prompt correctly.
I can’t find anything about Ghibli or Miyazaki’s names being on that restricted list.
Also if keyword blocking is the best they could muster, they were never serious about blocking certain styles.
From the article listed, a quote from ChatGPT:
Our policy restricts creating images in the style of artists, creative professionals, or studios whose latest work was created after 1912. Jim Lee’s work falls well after this cutoff date, hence the inability to generate an image based on his style
Right, but the point I’m trying to ask about is whether they’re treating Ghibli specially here. People are reacting as if OpenAI is thumbing its nose specifically at Miyazaki here, whereas the impression I’ve got is that they simply opened the floodgates and dropped restrictions on styling in general.
Style has never been covered by copyright to begin with, so any concerns they might have had about being sued over style would have always been erring on the side of caution. They may simply think that the legal environment has calmed down enough that they won’t be inundated with frivolous lawsuits any more.
I understand what you’re getting at, and this article was the best I could come up with. I think the real problem is that OpenAI is tight lipped about what they allow and don’t allow. As I said, I don’t personally use them, so I’m unfamiliar with if all restrictions are gone or if this is people doing the classic work-around-a-keyword filter. I have a friend who is exceptional about getting past their keyword filters in which he has done things he is definitely not supposed to be able to do.
I’ll see if I can get a hold of him later tonight, because he was generating some stuff in a Ghibli style in the last few days. I’ll ask if the keyword filter is still there and whether this is people just working around it, he would know better than I with first hand experience. Because I am having a hell of a time finding articles that actually detail what changed here.
I think we both want an answer to the same questions but the available writing on such questions is very limited, it seems.
They loosened moderation on style-based prompts. That’s the ‘real’ story. The End. But…
…some users on Reddit/X (hard to pin down exactly where, as these things go) made it a meme to ‘Ghibli-fy’ images because it is easy now (despite being trivially easy to do in ComfyUI for over a year) and then, in an attempt to monetize the meme/outrage, “”“news websites”“” started producing articles like this one were written using old quotes to imply that there is some sort of ongoing drama between OpenAI and Studio Ghibli.
It’s just manufactured drama built on Internet memes and outrage farming media sites.
Did they specifically allow “Ghibly style?” Or did they just loosen the restrictions on asking for styles in general, and Ghibly style just turned out to be the popular one that memes started snowballing around?
I don’t have specifics if they have allowed other styles to be used now, too. I don’t use this nonsense, but it’s clear that Ghibli was put front and center.
Yes, I read the article. But it doesn’t answer my question. Did OpenAI specifically enable Ghibli style, or did it remove the restrictions in general?
Everyone’s pulling out Miyazaki’s out-of-context quote about procedural animation and are interpreting this as some kind of personal attack against him in particular because of it, but unless OpenAI specifically made Ghibli style available without lifting restrictions on others I don’t see a reason to assume that.
Also, an article that calls X “The Nazi Network” is not exactly the most reliable source. This isn’t even about X.
That didn’t exactly look like animation. Looks like they trained an AI to control a humanoid figure in a virtual environment. It learned completely new and inhuman means of locomotion. Not very impressive from the technical angle, but the pitch about using it as a model for Zombie movement was clever.
You can use that for CG animation, of course. But those bi- and quadrupedal robots are also trained that way.
I feel the filmmakers manufactured some drama there. Knowing the real context of the quote makes it much more sensible.
https://bleedingcool.com/comics/chatgpt-wont-copy-artist-styles-including-jim-lee-frank-frazetta/
This suggests that all they’ve ever actually been doing is blocking keywords of artists names, and that it has always been trivial to get around such restrictions if you know how to prompt correctly.
I can’t find anything about Ghibli or Miyazaki’s names being on that restricted list.
Also if keyword blocking is the best they could muster, they were never serious about blocking certain styles.
From the article listed, a quote from ChatGPT:
Right, but the point I’m trying to ask about is whether they’re treating Ghibli specially here. People are reacting as if OpenAI is thumbing its nose specifically at Miyazaki here, whereas the impression I’ve got is that they simply opened the floodgates and dropped restrictions on styling in general.
Style has never been covered by copyright to begin with, so any concerns they might have had about being sued over style would have always been erring on the side of caution. They may simply think that the legal environment has calmed down enough that they won’t be inundated with frivolous lawsuits any more.
I understand what you’re getting at, and this article was the best I could come up with. I think the real problem is that OpenAI is tight lipped about what they allow and don’t allow. As I said, I don’t personally use them, so I’m unfamiliar with if all restrictions are gone or if this is people doing the classic work-around-a-keyword filter. I have a friend who is exceptional about getting past their keyword filters in which he has done things he is definitely not supposed to be able to do.
I’ll see if I can get a hold of him later tonight, because he was generating some stuff in a Ghibli style in the last few days. I’ll ask if the keyword filter is still there and whether this is people just working around it, he would know better than I with first hand experience. Because I am having a hell of a time finding articles that actually detail what changed here.
I think we both want an answer to the same questions but the available writing on such questions is very limited, it seems.
They loosened moderation on style-based prompts. That’s the ‘real’ story. The End. But…
…some users on Reddit/X (hard to pin down exactly where, as these things go) made it a meme to ‘Ghibli-fy’ images because it is easy now (despite being trivially easy to do in ComfyUI for over a year) and then, in an attempt to monetize the meme/outrage, “”“news websites”“” started producing articles like this one were written using old quotes to imply that there is some sort of ongoing drama between OpenAI and Studio Ghibli.
It’s just manufactured drama built on Internet memes and outrage farming media sites.