Tried to create a LoRA which gives women a more classic look rather than the current beauty standards. Based on 60's, 70's and 80's looks with wavey or curly hair and often with bangs. Triggerword not needed, but can have some effect.
I prefer to use a Flux.1 Schnell model, cause the Dev model gives a lot more generic "model-like" faces.
The used Flux.1 S model ArtFusion Flux v1.2 4-steps
https://civarchive.com/models/641214/artfusion-flux-4-steps
The used Flux.1 S workflow
https://civarchive.com/models/693450/flux1-s-workflow-with-lightning-upscale
Flux Dev (GGUF) examples
The used GGUF (Flux) Q8 Model
https://civarchive.com/models/647237?modelVersionId=724149
The used GGUF (Flux) workflow
https://civarchive.com/models/650990/gguf-flux-workflow-v20
Description
First release.
FAQ
Comments (8)
Thank you man. You did it again. Your models have the most beautiful classic faces. I am tired of the stereotypical Kylie Jenner face so your model here is a surprise but a welcome one! Thats exactly what i have been looking for!
That's exactly why I did it, it's always the same "Flux face" as I call it. Sometimes I also use the sideboob LoRA for this without wanting to show sideboob, just for more variation.
@JayNL Lol, i use sideboob too, not for the boobs but because of the beautiful faces in the last few example pictures. Also does good animated style images. "Same Flux face" so true. It looks the same all the time, only difference is that its sometimes asian and sometimes caucasion. Small lips or natural lips prompt is always ignored by Flux. I really have to make a lora of young Diane Lane (https://images.app.goo.gl/eLRa8dYkTFi5wnMD6) and young Susanna Hoffs of the Bangles.
What settings did you use for training? How many input Images we're there? I'm thinking about training my own style LoRA for 1.S
What kind of lora training algorithm was used? I tried tuning schnell using simpletuner but it doesnt seem to produce good results
Thank you for sharing LoRa.But I have a question, which is all the pictures I generate. The sense of AI is very strong. How should I set its parameters so that it reduces the sense of AI while maintaining its classical style?
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.