This lora was made with only ONE high definition image as a dataset. (240 repeats/10 epochs total 600 steps)
Another experiment to see how things turn out.

As expected the facial expressions doesn't change much due to only have 1 example.
But the clothing can be changed at will.
The background are really terrible, which I don't get since that's not trained in and should be done by FLUX itself. But it is what it is.

The reason I am testing FLUX loras with free resources or low image count is to find free/cheaper/easier ways for users who are not so knowledgable about Lora creation to be able to create models.

My previous high quality Fictional Model loras required paid subscriptions to AI services. And take around 6+ hours to make for just ONE lora. Which I upload and don't even earn the buzz back for.
This isn't really an option for new and inexperienced users.