늅늅 AI 첫 그림 과정 // 도움 받은 글 목록 있음

1. 갤 23 + 해상도 검색 ( 2340 × 1080 )

2. 스타뜨! (2/24)

3. 포기

4. 다시 시작 (2/25)

5. 챈에서 요령 검색

// 프롬프트만으로 깔보는 표정 만들기

// 대충: 프롬<>으로 묶으라는 글

// 존나 긴 네거티브는 쓰지마라

5-1 프롬 사이, 유기적 연결이 꽤 중요함을 알게 됨.

5-2 긴 네거티브 쓰지 말라고 했는데 씀. 병신

// [병합대회] MIX-Pro-V4_Beta // 에서 찾은 모델 사용

// 손가락 이렇게하니까 뭔가 더 잘그려지는거 같애 // 에서 지브리 LoRA 찾음.

첫 결과

masterpiece, best quality, < masterpiece, best quality >, < night sky, city lights, from below, (looking down:1.3), looking at viewer >, < 1girl, crossed arms, sitting, on couch >, < trench coat, pixie cut, gradient hair, black pants, white shirt, black gloves >, < (scoffing:1.3), open mouth, half-closed eyes >, (beautiful detailed glow: 1.2), white hair

Negative prompt: lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, artist name, lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, artist name, large breasts, thick thighs, huge ass, nude, (blue skin:1.1)

Steps: 22, Sampler: DPM++ 2M Karras, CFG scale: 9, Seed: 3673887415, Size: 300x650, Model hash: bced7c86c5, Clip skip: 2, AddNet Enabled: True, AddNet Module 1: LoRA, AddNet Model 1: studioGhibliStyle_offset(708c39069ba6), AddNet Weight A 1: 0.2, AddNet Weight B 1: 0.2

- 포즈는 포기한 그림에서 건진 유산

6. 대충 요령 검색

// i2i로 내가 뽑은 이미지 디테일 상승 시키는 법

1. 채택 2. 탈락

1번 설정값

< masterpiece, best quality >, (beautiful detailed glow: 1.2), (Increase quality:1.2), (smoother lighting :1.05), (increase cinematic lighting quality:0.9), 32K, < night sky, city lights, from below, (looking down:1.3), looking at viewer >, < 1girl, (crossed arms:1.2), sitting, on couch >, < trench coat, pixie cut, gradient hair, black pants, white shirt, black gloves >, < (scoffing:1.3), open mouth, half-closed eyes >, white hair
Negative prompt: (low quality, worst quality:1.4), (clothes removed:1.2), (bad anatomy:1.2), bad hands, bad_head, bad face

Steps: 22, Sampler: DPM++ 2M Karras, CFG scale: 7, Seed: 3562369743, Size: 300x650, Model hash: bced7c86c5, Denoising strength: 0.81, Clip skip: 2, Mask blur: 4

대충 퀄 올라간다 해서, 두 번 뽑음. 2번은 Seed만 바뀜. [ 3562369741 ]

2 번 그림으로 시도하다가 손이 어려워서 1번으로 우회함.

7. 얼굴 수정 요령 검색

// 얼굴 고치는 개인적인 팁

// 인페인팅의 질을 높여주는 무료 포토샵 사이트 Photopea

7-1 부가기능 - 리사이징, 296*648 -> 1184*2592

7-2 얼굴 자르고, 수정

< masterpiece, best quality >, < night sky, city lights, from below, (looking down:1.3), looking at viewer >, < 1girl, crossed arms, sitting, on couch >, < trench coat, pixie cut, gradient hair, black pants, white shirt, black gloves >, < (scoffing:1.3), open mouth, half-closed eyes >, (beautiful detailed glow: 1.2), white hair
Negative prompt: (bad quality, worst quality:1.4)
Steps: 22, Sampler: DPM++ 2M Karras, CFG scale: 8, Seed: 3562369741, Size: 512x512, Model hash: bced7c86c5, Denoising strength: 0.4, Clip skip: 2, Mask blur: 5, AddNet Enabled: True, AddNet Module 1: LoRA, AddNet Model 1: studioGhibliStyle_offset(708c39069ba6), AddNet Weight A 1: 0.2, AddNet Weight B 1: 0.2

- 네거 프롬, 간략하게 바꿈 + photopea에서 결과물 붙이기

< masterpiece, best quality >, < (trench coat:1.3), black pants, white coat >, (beautiful detailed glow: 1.2)
Negative prompt: (bad quality, worst quality:1.4), hands, finger, arm
Steps: 22, Sampler: DPM++ 2M Karras, CFG scale: 8, Seed: 3562369741, Size: 592x302, Model hash: bced7c86c5, Denoising strength: 0.4, Clip skip: 2, Mask blur: 5, AddNet Enabled: True, AddNet Module 1: LoRA, AddNet Model 1: studioGhibliStyle_offset(708c39069ba6), AddNet Weight A 1: 0.2, AddNet Weight B 1: 0.2

- 욕심 나서, 외투 옆 부분 수정 후, 인페인트

< masterpiece, best quality >, < crossed arms >, < trench coat, black shirt, white coat >, (beautiful detailed glow: 1.2)
Negative prompt: (low quality, worst quality:1.4), (clothes removed:1.2), bad anatomy, bad hands
Steps: 20, Sampler: DPM++ 2M Karras, CFG scale: 6, Seed: 2214454542, Size: 608x277, Model hash: bced7c86c5, Denoising strength: 0.45, Clip skip: 2, Mask blur: 4

- 다시 인페이트

마무리 작업 - 단순 합성