Type | |
Stats | 319 |
Reviews | (52) |
Published | Jan 3, 2024 |
Base Model | |
Training | Epochs: 6 |
Usage Tips | Clip Skip: 2 |
Trigger Words | janerodbd |
Hash | AutoV2 B16CCE86BB |
Weights
0.8-1 (also works in 0.5-0.6 with some models)
main tags
character
janerodbd, black eyes, lips, dirty face, high-waist-pants,
Looks
classic look
pinstripe jacket, pinstripe pants, shirt, high-waist pants, belt, black hair, single hair bun, dirty clothes, blood stained clothes, eyeshadow
pink shirt
pink shirt, tied shirt, sunglasses, summer hat, high-waist pants, jeans, outline heart, belt, purple hair, earrings,
complementary tags
This depends on the result you want to achieve but you may want to use some of these tags: latina, thick, dark skin, curvy, thighs, wide hips, large breasts, cleavage. Use at your discretion.
negative tags
the negative tag 'hands on hips' might be helpful sometimes
depending on the checkpoint you might want to use the tag 'asian' if it's giving too much of an Asian face and you don't want that effect
Recommendations and notes
This LoRA was trained with Anylora but when generating images I had way better results with other models.
I'm quite happy with how versatile it is, you can generate very realistic images and also can generate anime with it.
My favorite images are generated with the checkpoint and prompt style I use for my regular content (I can't share that info, sorry), but you will probably be able to use your favorite model and generate good images.
Depending on the checkpoint you'll have to play around adding or removing some of the complementary tags or the negative tags, as you might notice in the prompts of my example images, I had to add or remove some of them depending on the checkpoint.
Most of the images I created were generated with a weight of 0.9, so I recommend to use weights from 0.8 to 1. But I got some comments that depending on the prompts and model you might get good results with 0.5-0.6 too.
I'd appreciate your feedback on this LoRA since training is not my main thing but I'm really trying to learn and improve.