Type | |
Stats | 211 104 390 |
Reviews | (22) |
Published | Dec 5, 2024 |
Base Model | |
Training | Steps: 1,117 Epochs: 17 |
Usage Tips | Clip Skip: 1 Strength: 0.8 |
Trigger Words | RefinedBeauty_DA |
Hash | AutoV2 1F2E0950B4 |
About the Versions:
v1 Style: This version has a more pronounced artistic style and refined facial features. However, due to a smaller dataset and high training epochs, it is recommended primarily for portraits.
v2 Concept: An enhanced version, intended to improve upon the first. However, after extended testing, it proved to be unsuccessful.
v1 Concept: This was my first attempt – enough said. It was trained as a concept and served as an experimental starting point.
Trigger Word: RefinedBeauty_DA
This LoRA was created to enhance feminine facial features, adding softness and smoothness while maintaining expressiveness and uniqueness. The model addresses certain aspects of the base FLUX model, such as repetitive facial features and issues with split chins.
The LoRA is trained on a carefully selected dataset, featuring a variety of facial types and soft lighting effects that align with my own aesthetic vision of beauty. This adds a stylized effect to the images, making the faces more lively and visually pleasing.
My Settings:
LoRA Strength: 0.8-1.0, depending on the desired effect
Model: Flux.1 Dev GGUF Q8_0
Text Encoder: fp8
Sampler: Euler
Scheduler: Beta
Upscaler: Real-ESRGAN x4 Plus
There is no one-size-fits-all setting – experiment with the parameters to achieve the best results for your specific goals. Try different LoRA strength levels, as well as various stylistic settings and detail enhancers, to achieve your ideal effect.
This is my first LoRA, and I’m looking forward to seeing your works and hearing your feedback! Your impressions and results are incredibly valuable to me as I continue to refine and improve my models.