ghost
Sign In

Tiffany_Young_티파니_SNSD

52
568
2
Updated: Oct 5, 2024
celebrityfemalekpopkpop idolsnsd
20230603-3099978046-a photo of a woman , (tiffany young_1.2) , a closeup of woman face , rendered in octane , rendered in unreal engine , realistic.png
20230581-3363171375-a photo of a woman , (tiffany young_1.2) , a closeup of woman face , rendered in octane , rendered in unreal engine , realistic.png
20230549-2523596716-a photo of a woman , (tiffany young_1.3) , a closeup of woman face , rendered in octane , rendered in unreal engine , realistic.png
20230633-679287993-a photo of a woman , (tiffany young_1.2) , a closeup of woman face , rendered in octane , rendered in unreal engine , realistic.png
20230607-528298532-a photo of a woman , (tiffany young_1.2) , a closeup of woman face , rendered in octane , rendered in unreal engine , realistic.png
Download (567.64 MB)
Verified:
SafeTensor
Type
LoRA
Stats
568
Reviews
(53)
Published
May 28, 2023
Base Model
SD 1.5
Training
Epochs: 200
Trigger Words
tiffany young
tiffany_young
Hash
AutoV2
D0D83C6ADD
default creator card background decoration
zzaepharozz's Avatar
zzaepharozz
20230603-3099978046-a photo of a woman , (tiffany young_1.2) , a closeup of woman face , rendered in octane , rendered in unreal engine , realistic.png
20230581-3363171375-a photo of a woman , (tiffany young_1.2) , a closeup of woman face , rendered in octane , rendered in unreal engine , realistic.png
20230549-2523596716-a photo of a woman , (tiffany young_1.3) , a closeup of woman face , rendered in octane , rendered in unreal engine , realistic.png
20230633-679287993-a photo of a woman , (tiffany young_1.2) , a closeup of woman face , rendered in octane , rendered in unreal engine , realistic.png
20230607-528298532-a photo of a woman , (tiffany young_1.2) , a closeup of woman face , rendered in octane , rendered in unreal engine , realistic.png

Tiffany Young 티파니

whose smile shines brighter then gems

singer songwriter actress

vocalist of the kpop group SNSD - Girls Generation

LORA created from finetuned Henmix_Real

images created with Henmix_Real

Process:

  • 231 images gathered

  • upscaled images using swinIR

  • cropped to roughly 2x3 or 1x1 aspect ratio centered on subject

  • removed background text and other people using content aware fillers

  • normalized color to fix sepia or overly tinted images using neural filter colorize at 50-90%

  • normalized contrast , auto contrast then brightness/contrast adjustments when needed

  • captions generated using clip , then programmatically added trigger words

  • resized images down to 1024 max , training at 768 resolution

  • trained for 200 epochs with batch size 5 , at 3e-6 learning rate