Fairchilde Family Reunion Aesthetic (Dev vs Krea LoRA Training Comp Experiment)

Fix all your pictures so that everyone looks like they are related to Gen-X heart throb Lady Elaine Fairchilde. The degree which, from human relatives, to dwarves, to cursed puppets, is up to you.

Yes, I trained Lady Elaine previously and is my favorite LoRA, but it would be nice to have a little more flexibility and be less prone to being a legless puppet when I want to use her as an aesthetic and not a character.

This was an experiment/challenge for me- can I train a LoRA on Dev or Krea to work interchangeably with either model? Yes, but it took some work, and Krea just does not seem worth training specifically on (I think the trick is train on Dev but test on both. The Artsy Wrestle Factory I posted took this approach).

No trigger needed. Strength roughly correlates with familial resemblance up to uncanny puppet. May go a bit nuts with analog styles on Krea (often in a good way!). Strength ranges are roughly .5-1.2 for version D and .7-1 for version K.

I prefer version D.

Version differences:

·         D- dev trained at 512. Slightly better due to consistency. Versatile and easy to use. The wilder and darker of the two and will do some things cranked in Krea with more abstract styles. Also, the teats and bellies are bigger.

·         K- Krea trained at 1024. More realistic and softer but strength needs fidgeting with at times. Does soft cinematic styles spectacularly. More prone to artifacts in Krea. Works well with other character LoRAs. Does some real neat stuff with depth on dev at times.

Also note, if you ever want to train Krea, use short captions! Very short. Right click on the lora and open in Notepad if you want to see the captions. IMO, train the Dev lora a little differently and stick with one. I will likely be doing so.