同じ元画像からのimg2imgとLoopbackで一貫性のある肖像画を生成する
https://gyazo.com/a151df376a219e188bb7e364bfcf82efhttps://gyazo.com/6469cf03547b7b60ffeb5111788ea0c3
https://scrapbox.io/files/637ab0d9d88f8b002163de2d.webphttps://scrapbox.io/files/637ab0ef2055a1001d72b15f.webp
これはなんでも良い
テンプレート画像(元画像)も比較的何でも良い
ただし、これで構図が決定される
その分トークンが節約できる
ループバック回数はちょうど良いところを見極める必要がある
prompt
code:Arcane Diffusion
(black and white) digital portrait of _insert subject here_, 1920s fashion, chiaroscuro, detailed face, detailed clothes, cinematic lighting, reflections, rich shading, deep shadows, low-key lighting, high quality, by Guido Crepax and Tomer Hanuka and Yoji Shinkawa, from:arcane style:3 code:Negative
Negative prompt: (bad_prompt:0.8),comic panels, white outline, sketch, repetitive, cropped, bad illustration, kids drawing, long neck, totem pole, multiple jaws, (flat)
code:settings
46 - 48, Sampler: DPM++ 2M Karras, CFG scale: 12.5, Seed: {various}, Size: 512x768, Model hash: 318a302e, Denoising strength: 0.75, Denoising strength change factor: 1 (also --xformers).
その他promptまとめ