This version is trained on higher resolution images so the images are sharper and less mushy
Workflow
Dataset creation:
I've used gallery-dl (GitHub) to scrape over 700 Images of of different websites
I've removed duplicates with czkawka (GitHub)
I've manually removed all images smaller then 800px and images of low quality
I've used WD14 in kohya_ss (GitHub) installed with Stability Matrix ⭐ (GitHub)
Training:
I've put the files into a zip using 7zip (Download)
I've uploaded the zip to CivitAI for training
All left as default accept I've toggled shuffle the captions
This version is trained on higher resolution images so the images are sharper and less mushy
1. I diritti dei modelli ripubblicati appartengono ai creatori originali.
2. I creatori originali che desiderano reclamare il proprio modello devono contattare il personale di SeaArt AI tramite canali ufficiali. Clicca per rivendicare
