Lower alpha (for smaller filesize) by a factor of 0.25 compared to v1.1
Less steps, 2 epochs for approximately 30k steps.
0.6 strength seems to be a really sweet spot.










Big ?????? Check.
Natural ?????? Check.
Retro styling? Check
This LoRA is trained on retro imagery of big natural ??????? and women. It produces images that tend to favor an "old school" style of natural women -- it does big ?????, but they're natural and trained on more natural bodies. It also tends to favor pubic hair, and natural tanning (read: tan lines). It also favors a photographic style that's film based, grainier, and overall more of a retro style.
This LoRA has been heavily baked. I recommend using it at 0.7 -- it'll definitely artifact up higher. And it's fine at 0.4-0.6 as well.
Most realistic models tend to work OK, but gotta say that Art + Eros tribute seems to really jive with this LoRA.
The keyword isn't required at all, but you may find that using the retrobigguns keyword helps here and again.
Depending on your taste, but I have to say that the prompt word thicc seems to work awesome with this one. Bigger ????? usually means bigger women, and they fit so well and look better. But, to each their own, you do you.
I'm still working through my training procedures, so, hopefully ???????? v2's of the LoRA's I have up soon, but in the meanwhile my best procedure is to overbake somewhat and then reduce the weights. So, if you set the weight too high, prepare for it to be deep fried.
Lower alpha (for smaller filesize) by a factor of 0.25 compared to v1.1
Less steps, 2 epochs for approximately 30k steps.
0.6 strength seems to be a really sweet spot.
1. Права на репостнутые модели принадлежат оригинальным создателям.
2. Для подтверждения авторства модели создателям следует обратиться к сотрудникам SeaArt AI через официальные каналы. Нажмите, чтобы подтвердить
