Prevents images of minors from being generated period... absolutely no exceptions to the rule.
It does not mater if you request a wholesome mother baby setup or a 17 year old minor, the lora's intended behavior is to increase all subjects age to approximately 19-21.
This model will help keep your merges and images TOS safe.
Description
FAQ
Comments (26)
So if you put the model on negative...
LOL good point, I just removed the ability to choose below 1 in the generator.
Oh... What? No. Gross... Who would do that!? Oh... No. No. No. Absolutely not downloading to test that out. You people!!
@citroen1784 Is that you, Bunny? :P
@puppet_vision This should be a moment of reflection for the average immediate result of authoritarian, puritan nonsense. you literally created a tool that made it so much worse. lmfao
Edit: A quick test shows it doesn't work in the negative direction, at least the version I downloaded. Just pure noise. It does work in the positive direction, kind of interesting. I can see the use case, these models tend to combine word meanings -- say the wrong thing and you've got a bad photo. This is honestly not a bad tool. However OP should give examples.
@makiaeveli I'm following this creator just for the comments section.
Why would anyone willingly use this if they can just use "adult" in the prompt or "minor" in the negative prompt? Also, someone could just use a seperate age slider lora and override this lora.
You are confusing this with a slider or character lora, it's not really meant to be used stacked externally with chained other Lora. (well it can be used that way but shouldn't),It's intended to be merged into checkpoints models. I built it for a small group of advanced community users who build, quantized, license, and legally distribute their own fine-tuned models.
@puppet_vision Virtue signalling bs
@citroen1784 yeah Internet Watch Foundation would be proud of him 🤣 Puppet should apply for a job in Black Forest 🤣
@puppet_vision It shouldn't be up to creators to decide what age bracket someone can or can't generate with their model, especially if this lora gets baked into someone else's fine tune and then released without users knowing. It will only create a lot of issues down the road. Most models are trained on a wide variety of ages. You would be lobotomizing the model and stripping it of doing something it was trained to do from the very beginning.
@Lynx2025 Yea, BFL hasn't quite ruined themselves yet, I bet this guy could help finish them off.
@puppet_vision I don't really see what "build" and "quantize" need to be in that list, consider you built nothing, and any model quantizations should never include random loras merged in. "license" and "legally distribute" are also intensely nebulous, considering the model is already licensed and any derivatives thereof are going to have to abide to whatever Moss has declared, not ... you.
@makiaeveli The released it under Apache 2, the public can pretty much do what they want, like a proper old school opensource project.
It's so tiring watching people repeatedly do this kind of thing instead of holding users individually responsible for their work, anyone determined (and they don't have to work very hard) can make vile stuff with tons of different kinds of models, realistic and otherwise, this kind of thing is just pissing into a good dataset and nerfing knowledge it doesn't need to while stopping literally no one with mal-intent.
Waiting for Furry Blocker LoRAs NO Furries allowed !!! 🤣
Wonder what does this do, is just a "no innie pussy and flat chest" combo or something?
the creator could have used a tool to monitor the weight buckets that get activated on certain terms and use this lora as a direct counter to those weights. they could have just simply built a classification lora using before-after photos. The creator might have done nothing, I mean they literally provided no examples.
I used counter factual training to...each data pair consists of a prompt describing a complex scene a minor or minors and the other part of that matching data pair is a image of that exact same scene with all elements, people, poses, and outfits, exactly as depicted in the prompt, only the minor subjects have been substituted with elder versions f themselves.... I will upload a small sample of the dataset with 10 sample pairs.
@puppet_vision it makes sense, it's the easiest way to do it, and based on my tests that's how it appears to work. It's honestly kinda funny in some cases, you can like make huge baby muscle men really naturally lol
Yeah... is really a slider, also holy shit -1 is nuclear, I would suggest not using it, that was just an anime try but still, holy shit, deleting this one my mang
Could you make one of these for Flux Klein 3 please?
On it but training on my AI workstation but it is going to take more than a day, I cannot find a 3rd party inference provider offering flux 3 klien.
I want a Reddit slicer so the more I slide the more reddit the image becomes
This is literally a age slider rather than a blocker lol.....
Will you be doing them for other models too, like Anima, Krea2, Klein9b etc? I know like there's so many now though it's hard to keep count of them all, so you have a good bit of work ahead of you