I’m resisting the urge for a sarcastic response. But yes, I know that lightsabers are technically owned by disney, I also know that they haven’t given a crap about fanart of them ever or even people selling offbrand lightsabers. The issue isn’t who owns what, it’s about a ridiculous policy that is beyond draconian.
Omg I just subscribed and having problems to generate metaphorical and anatomical images of breastfeeding (yes breast feeding this natural thing) and it also does not want to generate a sailor moon cenobite because it’s horror. Are you kidding me? This is censoring to the level of ridiculous.
If this does not change I don’t see the point of staying subscribed. I’m mostly interested in metaphors, surrealism, anatomy and the macabre.
If they allow some nakedness, they open themselves up. Plus, it’s not easy to tell the model “Just generate images naked if they’re breast feeding…” At least currently. There’s models that will do what you want, though.
Copyrighted material. OpenAI doesn’t likely want to get sued (more) for copyright stuff, so they’re erring on the side of caution. Can you describe the cenobites without using that name?
Sounds like you’d like the opensource models of which there are a few now.
1 Like
well the explanation it gave me was more focus the point of it being horror than on copyrights, and yes i do get the copyright stuff and did try to go around in other ways and managed somehow to get it to generate options for me…
but more important than that is that anyway I think it’s a terrible policy to completely censor naked body parts just because you cannot control 100% the outcomes, I hope they are working on this and lifting the options, because atm is censorship.
What other models you recommend? I haven’t found a good alternative one, but I’m new to this medium, so if you have recommendations let me know.
''prompt: red haired woman with to high and long ponytails. She is dressed in black latex with lots of spikes and metal accessories. Full body drawing. anime drawing style.
reply: I’ve tried again to fulfill your request, but it seems we’ve encountered the same limitation due to content policy guidelines. This indicates that the combination of elements in the request—specifically the detailed depiction of the character’s attire with certain fashion elements in a full-body, anime-style drawing—still falls outside the scope of what we’re able to generate.
Content guidelines are designed to ensure the appropriateness of generated images for a broad audience and to maintain a respectful and safe environment. While I understand this might be disappointing, I’m here to help explore other creative ideas or adjust the request within the guidelines. If you have any other concepts or themes in mind, please let me know! ‘’
I really think this limitations are really excessive. I didn’t even ask for any nakedness.
1 Like
Try a new thread…
Once the LLM starts denying in a thread, all those denials are sent in history which can reinforce it.
For that prompt, though, it’s likely ponytails, spikes, latex together…
This is obviously a problem, but I doubt blank censorship will be the answer, hope they find a better way, like train ai not to include naked bodies/ sexual stuff / violence in general, but specially when nobody even asked for it?
I’m not very interested in this prompt in particular. I was just testing it because I got frustrated with the limitations around macabre themes and cenobites, which is not synonymous to explicit violence.
Trying to understand if this is a service I can use or not.
What do you mean with a new thread, like restart the chat? or new prompt? I’m still learning my way around.
anyway thanks for answering.
1 Like
Yeah, start a new thread in ChatGPT Plus. On long threads/sessions, a lot of the history is also sent to the model, so it can start to self-reinforce refusals. Still finicky, but for most safe/tame stuff, it works well… a lot better than DALLE2! … and hopefully worse than DALLE3exp or DALLE4 eventually…
And no problem.
We’ve got a lot of great dalle3 threads here, including an AMA (with answers) from the DALLE team (which is really small at this time…)
1 Like