Embeddings
Changelog
v3.0
Less inclined to produce naked body.
Trained on Pony Diffusion V6 XL.
v2.0
More consistent facial features.
Trained on Pony Diffusion V6 XL.
v2.0-alpha
Failed experiment - definetely under trained, however I kinda like the visuals.
Trained on Pony Diffusion V6 XL.
Notes
My first LoRA (I'm super noob)!
Since this is the first half-decent result that I got, I thought to share it :)
It's probably overbaked, I still can't choose the "perfect" epoch accurately :(
Trained with OneTrainer using additional embeddings, so both the LoRA AND the embeddings are required to properly work.
The character is a sinous vampire girl with red eyes and black ponytail, wearing a tight black leather outfit with gothic and victorian elements.
I tried to train character/outfit separately but without much success - there is some bleeding over between concepts.
Also, I trained the character with naked images thinking that it would be more flexible when swapping outfit, but this kinda backfired.. It's causing some generations to be NSFW - in case this happens try setting completely nude in the negative prompt.
Trained on Vendo Pony Realistic
Samples images generated with:
refiner: DatAss Rev 3
If anyone has any advice to give, it'd be very welcome!
Trigger words
Character: m0r4n4
Outfit: g0th1c80dy5u1t
Both: m0r4n4 wearing g0th1c80dy5u1t
Description
First (clumsy) attempt at training a Flux Dev LoRA.
There is no embedding, it is a simple LoRA.
I can't seem to be able to prevent the grid-like pattern, if someone has some advice in regard please DM!