JH's samdoesarts : v5: jhsSamdoesarts_v5.ckpt

For v5, data is split :

• face closeups - Reg : a subset of ffhq by nvidia
• portraits - Reg : women digital portraits and photos generated by sd1.5
• zoomed out / landscapes - Reg : generated by sd1.5 using BLIP captions from training data

Original artwork and all credits go to Sam Yang (samdoesarts) 🎨

If you liked the models, join my 🎥 insta, or let's talk on discord: JH#7326

Training info :

• Models finetuned using kane's repo : https://github.com/kanewallmann/Dreambooth-Stable-Diffusion
• 622 final images captioned using BLIP
• Using 3090 on runpod.io