Better and more natural looking faces | Flux.D
Details
Download Files
Model description
The purpose of the LoRA is to first of all give you more variety when it comes to faces, it also improves quality (adds texture/pores), and hopefully makes the skin look less plasticky. I say hopefully because I haven't had time to test this LoRA extensively, but so far the results look decent.
Recommended strength: 1.0 (but experiment with it, since I haven't tested much)
The LoRA has a trigger word "face", but it works without it too. The prompts are included in the uploaded pictures so you can get an idea of how to use it.
The LoRA was trained on ~2000 images, a small amount of those pictures were NSFW (not by choice, just happens with such big datasets) , so in rare occasions you might get explicit generations.
I hope you enjoy this LoRA, feel free to post your creations and to give me feedback!
Comparisons:
