Verified: 10 months ago
SafeTensor
Type | |
Stats | 102 0 |
Reviews | (2) |
Published | Jun 21, 2024 |
Base Model | |
Training | Steps: 4,000 Epochs: 4 |
Usage Tips | Clip Skip: 1 |
Trigger Words | Al1a 1girl, solo |
Hash | AutoV2 07091EB5ED |
LoRA based on the best images I could find of Bollywood Actress Alia Bhatt.
~` Trained on` ~
Photon 50 images, 20 steps/image, 3 epochs using kohya-ss-GUI
~` PLEASE READ `~
Use Detail Tweaker as used in the sample prompt below !!
~` How to Use `~
Prompt (Use Lora at 0.8-1.0 for best results):
<lora:Al1a:1>, Al1a,, smiling, perfect small nose, nose, solo, 1girl, looking at camera, realistic, photorealistic, long black hair, perfect black eyes, highly detailed face, very high quality face
Negative prompt I recommend (for SFW):
fcNeg-neg, ng_deepnegative_v1_75t, negative_hand-neg, verybadimagenegative_v1.3, Unspeakable-Horrors-Composition-4v, big nose, large nose, multiple girls, nsfw, nude, naked, nsfw, nudity, deformed body, mutilated body, child, ((anime, art, drawing, painting, cartoon, jewelry)), child, young, big eyes, very big eyes, large eyes, big jaw, big teeth, eyes wide apart, wide eyes, ugly, bad eyes, fat face, fat cheeks, earrings, ear rings, wide jaw, wide face, old, cute
Other settings:
Sampler: DPM++ 2M Karras
Sampling Steps: 30
CFG: 3.5-7 (VERY IMPORTANT !!!!!!!!)
Clip-Skip: 1
PS: Alia's face was a VERY tough to make the AI learn, it learnt alot but didn't learn much, not my best LoRA.