This LoRA was trained on a custom dataset of diverse photos curated based on my personal preferences. The primary goal is to enhance facial details.
🌟 Key Features:
Best suited for: Realism and Semi-realism.
Anime Models: Works with anime models for face modification, but keep in mind that this will bias the overall style toward realism.
Pose-friendly: Has almost no effect on poses.
Recommended Weight: 0.5 - 1.0 (adjust based on the checkpoint, but it handles most models perfectly within this range).
Trigger word: Not required
💡 Recommendations:
I highly recommend using Face ADetailer for the best possible results.
For the preview images, I used this workflow.
⚙️ Technical Details:
Dataset: 45 images
Optimizer: adamw8bit
Learning Rate (lr): 0.0001
Batch Size: 2
Steps: 2400
Linear Rank (dim): 32
Conv Rank (alpha): 16