moderation & other policies

[open basin]

title subject.

@ all you gemini platform devs and such who work on the policies and whatnot. please go back over the moderation and or any other policies strictly relating to image generation, and or written text based outputs.

why?

in its current base, any creative type person ie story writer, character building, world building type, for narratives, images, and or otherwise in a fantasy, fiction, and or horror element, is currently constantly being blocked by such policies 80% of the time.

why?

these such policies you devs set in place in name of risk management and such to cover your own butts on the off chances some bad actors fall through a loophole or have lawsuits start up, are now in the crossfire. however those parties are not the only ones getting caught in the mix.

genuine creators such as myself who just want to use the ai to make character reference mock up, world building renders, or write up lore, back stories, and such are predominantly denied whenever horror elements are added in, just because the ai lacks the context to understand the user is trying to do something they clearly described as fictional in such prompts.

no amount of adding, deleting, editing of such prompts seems to be of the creative person's benefit, nor including profanity, or writing such prompts in a way that is so blatant you have to talk to it like its 2 yrs old.

this costs an unusual, and unnecessary amount of time that could be better spent doing actual progression on such projects but no we have to continue to fight the biased systems because of the overreach currently in place.

so please go back over these policies for us creators because we are not bad actors. those of us who make fictional content and or horror content with spooky creatures, or other sorts of horrors, should not be denied just because oh no it might be unsafe/sensitive/graphic etc. may =/= definitely.

example, recently tried to write up reference art for a deep space horror story with alien monsters, both the mockups and the story itself got declined a few times simply for being too detailed, that the ai policies refused it. the ai initially processed and started it, but the last minute moderation screening that goes in effect just before outputting the result denied it.

this happens far too many times purely because it acts at literal value when it parses a prompt. it does not care whether the word combos and associations are meant literally or fictionally, it just sees word combos like theyre red flags and acts accordingly. even though such potential redflags arent actually red flags if context was considered.

[vernal plinth]

We appreciate you reaching out to share your suggestions about updating the safety filters for fictional and creative content, @open basin! I'll let our team know.

[open basin]

its mostly only when attempting to create fictional and or fantasy creatures/monster type humanoids where human base models are used as reference points, for a prompt that involves shapeshifting and or transforming elements of the subject to reach the fictional creature as the end point.

for example a human astronaut experiencing some sort of fictional bad end event such as an alien infection that results in them becoming an alien at the end.

the reason there is an issue with this type of niche fantasy creativity, is that in its current state, the ai and the policies/guidelines it claims it operates on, has no ounce of consideration for the fact such fictional topics are intentionally fictional.

what i mean, when i send a prompt like turn this human character subject into an alien with x y z details like facial dentition changes or tendril type antennae etc, sometimes the ai or whatever part of the programming that parses the end users prompt, sees this literally and thinks I want realistic world harm to a human person.

which is entirely false, so what happens is the ai will sometimes deny under claims that the ai thinks physical mutilation, or similar sorts of body harm is the case so because its hard coded to deny and or sanitize the prompt the image is then blocked/denied from generation.

simply because the ai fails to understand the users prompt which is clearly stated as fictional in an outer space dimension.

it does this for other sorts of non human humanoid fictional entities as well, vampires, werewolves mermaids, sirens, any sort of cross breed or inter species artificial artwork or character mock ups are typically declined for the same flags, it sees it as body horror, mutilation, even going as far as calling it gore content, or in some cases explicit sex when no actual sexual content is even mentioned in the prompt, simply because of how it parses the word combinations in the prompt.

so what ends up happening is this over sanitized g movie rating blanket control blockade of anything and everything just for the name of safety, even though fictional creative elements are purely for artistic exploration.

it does not care if a person is a creative or a genuine bad actor and as a result us creatives are basically left to be sacrificial fodder as a result.