- This topic has 11 replies, 4 voices, and was last updated 1 year ago by
.
-
Posts
-
It'd take forever. The posing is the thing that really slows down my Daz scenes, because sometimes there's so much fussing about with the controls to get it right.
That looks great, though. Did it take long to get the desired results?
Those were the first images that came out because I was only testing the lora. I used to love putting scenes together in DS but I don't have the patience for that anymore now I'm old. Most of the stuff I do ATM are semi-portraits or nostalgia.
The Hong Kong one is not mine, it's from Civitai and I was just giving it a try out. I don't have a PC good enough to train Flux loras, at least not in under 6-8 hours.
@frank22 Yeah, training LORAs is rough, that's why I avoid it unless I really want something specific.
I only trained one with CivitAI (I had a lot of points to spare), by I didn't really know what I was doing, so it wasn't that great.
Yep I really miss lora training, it was fun. For SD1.5 it only took about 40mins to train a lora with my 4070 but I'm not prepared to wait 6-8 hours for a Flux one. I'm also worried about having my GPU running at 100% for that long. What I really need is a 5090 - I might have to visit my wealthy grandma with some delicious cupcakes (+ cyanide).
@frank22 Yeah, I also want the 5090, but it is really expensive and power hungry. My 3090 needs 300W, the 5090 600W, so that also needs a new PSU.
Ah, well, for now I am not training LORAs anyway, Kontext is too fun as is, plus with the nunchaku nodes it runs at x3 speed (it also works for flux).
- You must be logged in to reply to this topic.
