Lora trained for 20 epochs on synthetic/semi-synthetic data featuring a male human and a female robot in romantic settings. The style of the training data is burned in, that's why it looks so colorful but I don't mind it, I find it endearing. Training data consists of Dall-E 3, Midjourney and a couple of SD gens, some were inpainted or edited in Photoshop.
Trained on Anything v5, tested on Counterfeit and Anythingelse v4.
Description
FAQ
Comments (5)
For those looking to create the perfect "human + robot" couple, it might be a good idea to start with the images provided by the author for testing and then gradually adjust them in the desired direction.
(For those, like me, who are not quite clear on how to modify the prompts.)
What model do you use it with?
I use a hybrid checkpoint based on AnyLoRA.
Just ran some additional tests, and I won't encounter this issue if I generate images following the prompts you shared.
It seems to be my issue, so no worries,the only thing that needs fixing is my prompt.
I've updated my comment to make it appear more polite and friendly.
Thank you for generously sharing such an interesting Lora.
@Junbegun thank you for your feedback!
any love for lesbians? :(