While even the default SD 1.5 generates very nice trees - I couldn't find a way to generate exactly the style I was looking for. Hence I created this LoRA - it's nothing sophisticated, just a simple LoRA trained on 50 high quality photos taken by my girlfriend in the Rogowo Arboretum in late October. Trees were yellow and red, sky was blue, sun was shining.
Use it to generate nice delicate colorful trees or leaves. While it was trained on the images of trees and leaves only - it's also good for backgrounds too for all your mythical animals, witches, wood elves or other chineese cartoon underage girls with overgrown breasts.
rogowoarboretum is the trigger word, use additional keywords like "red leaves", "yellow leaves", "yellow trees", "red trees", "pink flowers", "red berries", "red fruit", "park bench", "woods", "forest", "close up" and so on.
LoRA weight of 0.8 or lower.
Description
FAQ
Comments (8)
"underaged girls with oversized breasts" thats so sus lmao I get it's probably taking the piss but that alone makes me not want to download this model
Well... that was me being a bit sarcastic... most of the things people here seem contain them after all which not only seems a bit boring but also doesn't show the potential this tech has... O_o
All those people want it for is to break the law with their disgusting obsession with children we need more male based LoRas to flood out all the shit and so I can actually make more male characters because I feel very boxed into mostly just female besides a few decently made male ones I have most of them they train it on like 10-20 pictures call it a day
@Merkuriij yes, the AI bias is now more real than ever. Make no mistake. Even when I tried LoRA trained on my own photos - adding "beautiful" somewhere in the prompt automatically gave me feminine facial features. Using the base Stability AI models and avoiding such adjectives generally helps.
Yay! 2 other people who are tired of seeing underage girls with oversized breasts. The amazing objects, creatures and landscapes Stable Diffusion can produce continue to astound me. So does this focus on copying the same over-sexualized young girl images. "Oh look it's like a camera, but I don't need to get near a real girl, and I can make her tiddies bigger!" It's very sad seeing these people rule the world. And how trained some models are at this. People use simple prompts for girls and get arched-back bimbos in too-tight clothing. I ask for a "hookah smoking caterpillar" and get 3 naked girls in a row. On Lexica a "Magic Space Ape" produces a pretty girl face.
I could rant for hours on this, but I did successfully train my first LoRA. Well, I didn't run out of VRAM and the model is recognized. I'd have been training models and LoRA like crazy if I didn't have only a 1080 Ti and few funds. I guess at some point I need to make a Patreon or GoFundMe or whatever to get some $ for a new GPU. Sadly if it was to create more little girls with torpedo tits, I'd probably raise a lot of $.
@HanJammer I agree, but it's sad because some of the new models are so much better at other things. I end up using them and a lot of negatives like "girl, lady, woman, female, face", etc. Sad that I've looked at many prompts where they just asked for "hero" or "cyberpunk" or even "portrait" and a girl was assumed.
@parallelepipedon I have the PC power to make LoRas I just haveno idea what I am doing
It was such obvious sarcasm that to not understand it was to admit one's own dementia.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.