This is my first flux trained model. It uses the same dataset as my Wolfcut DORA models. It seems to run well with 0.4 - 1.1 strength.
Description
Version 1 - Lets see how it goes
FAQ
Comments (4)
Works pretty well! How did you caption it to train for the hair?
This one was a lazy one, I left the captioning the same as my SDXL version. The first 3 tags are 'wolfcut', 'hair', hairstyle' which should generally capture the main parts. Then I describe the subject and a general one for the background.
I'll post the dataset later today for you. But I want to reception using natural language. :)
@Evivrus Thank you! I haven't tried with Flux yet but on SDXL I always had the problem that it learned the face too much. I inpainted or face swapped it out afterwards.
@LiaZ1 Ah yeah! You need to make sure you have a diverse dataset, if you have the same person multiple times you risk learning the face. You can censor the face and tag it with "face censored" that is also a good way to get around it :)
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.