Verified: a year ago
SafeTensor
Type | |
Stats | 140 0 |
Reviews | (13) |
Published | Apr 17, 2024 |
Base Model | |
Training | Steps: 1,475 Epochs: 20 |
Usage Tips | Clip Skip: 1 Strength: 1 |
Trigger Words | kelly brook kelly |
Hash | AutoV2 55EFD5EA3A |
Lora of the celebrity Kelly Brook, trained on 300 images at a variety of aspect ratios, settings and ages, so is fairly versatile.
The model was specifically given images without all the airbrushing found in many photoshoots to generate more realistic looking skin and helps eliminate the blurring effect. It does also respond mildly to user input ages if you want to focus on her earlier or later looks.
Can do close-up or zoomed out shots, just add necessary prompts like 'close-up', 'full body portrait'
Recommended lora weight: 0.5-1.
If using Hi Res fix, I recommend (for all realistic character loras in general) adding the prompt ('big eyes') as the smoothing tends to shrink them.