This man needs no introduction. You know him, we love him.
Support me on KO-FI so I can afford more rendertime, thank you :)
Or leave a rating, both would be appreciated.
And this might just be the best working LoRA I've created so far. The prompts I've tried were just a few minutes of experimentation, and literally 8/10 worked on the first shot.
It just works.
The training was done with the kohya LoRA model, so it requires the LoRa extension for Automatic1111 to function.
The model was trained against RealisticVision 1.3 but produces good results with artEros, hasdX and uberrealisticDreamy as well. Please read the rest of the description for more details.
I've used 50 Steps, CFG of 7 and a model strength of 0,9 for all images, the prompts were standard short model photoshoots, see the examples for the actual prompts used for each image.
One recommendation, try to avoid the word "suit". It was almost impossible to find pictures of this man where he doesn't wear a suit, and so the dataset was heavily biased towards suit pictures. Use the word "armor" or "cosplay" instead to put him in different outfits.