Trigger Word: pl4nt_91rl_l0v3

Weight: 1.0 is too high (maybe 0.5 to 0.8?)

Lol.

This is my first LoRA. That's about to be very evident.

Or rather, this is the first one I am releasing. My first two didn't work out well. They were a bit wacky.

This one is even wackier.

Look, this one is weird. It's not going to do what you want usually. But... it seems kind of unique.

It was trained on images I created from Bing Image Creator. (a few of the images I modified to correct some stuff) Those images, the prompts I used (which can't be replicated due to how that creator works) and the Kohya_ss settings I used to create this are included in the zip file.

The magic word is "pl4nt_91rl_l0v3". I read that if you don't use an actual word, you don't impact the prompt. So you could add only that and it will generate something. (maybe add things to make it look realistic if you want and add a negative prompt)

I didn't know how to create a Lora, and I still don't, but it gives some wacky imagery that is kind of cool. Basically, I wanted a woman made of roots in a plant pot to be holding a human woman. They stare lovingly into each other's eyes. Don't you judge them!

That of course didn't happen. Well, usually. Maybe I haven't found the right prompt yet. Good luck!

I should have taken out some of the images in here, because they generate some really different things sometimes, but I wanted to give it a purposeful randomness at times for the heck of it. And I would say you'll get some random stuff. Whether because of what I did leaving in images that were different from the concept, or because of AI wizardry, I don't know.

In case you weren't paying attention, I DON'T KNOW WHAT I'M DOING!

You can use this model however you want, including for commercial purposes, without credit. You can use the training images to make another Lora and don't have to credit me at all. Just have fun!

This LoRA was trained on Realistic Vision V6.0 B1 [Pruned Model fp16 (1.99 GB)]:

https://civarchive.com/models/4201/realistic-vision-v60-b1?modelVersionId=130072

As of posting this, the button says "V5.1 (VAE)" there but the file downloads as: realisticVisionV60B1_v51VAE.safetensors

I would like to see any cool images posted here. I would love to see what prompts and models work best.

I don't know what weight to use. I say 0.5 to 0.8 maybe, but sometimes some of the ones at 0.4 are good for the woman to be only slightly a plant. Depends on the model. Also, a lot of times one of the women is not human. You have to try to negate that.

And sometimes, more like often, things are just crazy.

I can already see where I messed up some training words. Oh well. I added "1girl" and "2girls" to every caption. Also, I didn't know where the setting for "Clip Skip" was until now. I had it set at 1. Maybe I should have done 2.