I honestly regret making this LoRA. I would unpublish this (like the rest of the artist-style LoRAs I did), but I made it an early access model when I published it a while back, so I can't unpublish it. Call it a moral crisis, but yeah, I think that actual, individual artists should be supported (I consider LoRA making a fun craft). I made a few style LoRA back when I was newer and didn't really grasp the scope/"damage" style LoRA can cause to individuals. Anyway, rant aside, you can find Hara's Pixiv page at: https://www.pixiv.net/en/users/825270, so please consider supporting them! I will wipe my hands of individual artist-style LoRA moving forward (I consider individual anime characters and anime art styles "fine" to train a LoRA for, since the studios are funded and they're more commercial and publicly facing works). Anyway, yeah, I'm rambling, but that's my stance!
I'm probably playing mental gymnastics, so short version: I'm trying not to trample on individual people whose livelihoods may be at stake.
➼Like what I do? Consider throwing me a few dollars at https://ko-fi.com/dandysombraes or commissioning me https://www.fiverr.com/metalchromex/make-you-a-stable-diffusion-anime-character-model so I can continue making cool crap for you all to enjoy for free!
Note: I'd suggest having something like (uncensored, no mosaic censoring) in your prompt and (mosaic censoring) in your negative prompt to avoid forced mosaic censorship.
Description
FAQ
Comments (9)
I love this artist art style, is this created on civitai ?, how did you compressed 900+ img datasets to 50 MB (cannot exceed 50MB)
So, I trained this model with a DIMM/Alpha of 32/16. A DIMM/Alpha 8/4 would get you the 50 MB-sized file. I usually do 16/8 and 32/16, as they seem to get me the best results (8/4 generally works fine, but I've seen slight improvements most of the time with a larger file size).
This is probably one of my favorite models for creating my blended styles, but out of respect for your convictions I will stop using it and find another.
Yeah, I don't want to come across as 'holier than thou' (I mean, all this tech is basically visual plunder/remixes to begin with), but as someone who has often struggled with money/finances in the past (and still has college debt to pay off), I've become more aware that I may have indirectly caused some damage to freelance artists. I'm not going to say people can't use this model, of course, but it's more me trying to course-correct and be "the better person" here, I suppose.
what happened?
I had a moral crisis of sorts lololol
@metalchromex what do you mean by that ? Something happened to the artist?
@Dineko the way that ai loras are trained affects directly to the artist that was used for training, due to not receiving permission, nor any retribution for their work used as a database. so yeah, it harms artists, heads up to metalchromex for being aware of it. Respect!!!
@Nsfwmann Yeah, that's correct. Sorry, Dineko, forgot to respond. I was kind of naive/didn't really understand the full scope at the time, and it only really started to click semi-recently. Hence why I'm only sticking with publicly available/commercial works moving forward that aren't done by just one person. I know that probably still makes me a hypocrite, but I prefer to champion the individual over the collective, if that makes sense.