V1.5: more flexibility and high rank, so you can do a merge with mostly common porn loras but at cost with quality issues.
Make nudism without biased faces.
this lora is not capable of sex and closeup.
Description
FAQ
Comments (10)
how many steps did you train up to?. I'm currently at 6000 and vagina still not good.
3250, i did a lot of stops changing datasets and parameters.
Unsurprisingly literally everyone else is using defaults and here comes LuisaP with dat 20mb LoRA
fuck yes
thanks for being you and thanks for the LoRA
I was thinking "160MB loras aren't bad compared to the sizes for some previous models) and then someone who knows what they are doing shows only 20MB is needed!
@DrStalker - we have 6mb flux LoRAs that work perfectly though, right alongside 600mb ones. And z is just 6b so I can imagine LoRAs smaller than 20mb being effective.
@DrStalker It depends on the style and dataset. Sometimes 8 dimensions are enough, other 32, oher 64...
i need pubic hair,
Then use ZImageTurbo as it is.
stop shaving!
Is this a personal confession or....