V7 WE DID IT!
The results are very consistent.
Pro tips: Use detailed prompts like:
<lora:Twinks_Legs_Up_V7:1> artistic intent to portray the subject in a natural and unobstructed manner, with light skin and straight brown hair, wearing only white ankle socks and a thin silver chain necklace. The man's nude body is lean and muscular, adding a tactile element to the scene. The setting appears to be indoors, with slightly parted lips and direct eye contact with the viewer. His body is slender with a toned physique, lying on a bed with his legs spread wide, and a small plant. The walls are painted a neutral color, with his hands resting on his lower legs.
The model still has a penis deficiency use img2img with androflux or twinkx to fix that if needed
Square images work the best!
Good images at least in my tests: 8/10
If you like the work of Human Diffusion consider tipping Buzz or Posting Pictures with our LoRAs
Description
Best Results:
1320x1600
Euler, Scheduler: Simple
In General Square images
The LoRA is still trash. I have no idea how flux interprets positions so if somebody finds settings that give better. WILL PRODUCE NIGHTMARE FUEL
If you get a good result without a penis use Androflux to inpaint the penis after
On the bright side we dropped the nightmare fuel from 5/10 to 3-4/10 so profit!
FAQ
Comments (5)
Your first example image is still the best looking penis I've seen on a flux lora, so you're making great progress.
Hehe thanks, the kudos go to Androflux though, my Lora has no idea of penises since it mainly has seen balls. The dataset has legs up-twink pictures which usually have the penis hidden due to the pose. Tried to use a combo of my LoRA and Androflux to see if Androflux could save my LoRA, It does work 1/100 generations.
But thanks again for the positivity!
Do you use Civitais trainer or train locally?
I use either depending on what i wanna achieve. The Kevin Spacey, Jordan Peterson LoRAs where Civitai Trainer. Taylor, ChasteFlux and this one Legs Up is trained locally.
In general normally if my LoRAs have more than 1k steps it’s local!
@HumanDiffusion Great, thank you for the reply! I have a 3090 arriving very soon and would like to locally train a few flux models. If you have any tips or parameters you would like to share I wouldn't complain! If not that's also fine.
We will bend Flux to our will XD.