Dansori_EQ/docs/IMAGE_GEN_HAIR.md

IMAGE-GEN RIGGING · 헤어 라이브러리 (머리 + 표정)

역할: 어느 바디에나 목(neck) 앵커로 합성되는 머리 라이브러리. 머리 = 얼굴 + 헤어(모양). 헤드폰 없음(헤드폰·머리띠는 악세서리로 얹음). 색상은 이미지로 만들지 않고 코드 틴트로 처리. 조립/색상 규칙: IMAGE_GEN_RIGGING_GUIDE.md. 항상 sori_sheet.png 첨부.

★ 규칙

• 머리 모양 4종: short(단발)·long(긴 생머리)·waveS(짧은 웨이브)·waveL(긴 웨이브).
• 각 모양마다 ① 머리(기본 neutral) · ② 표정 프레임 세트 · ③ hairmask(머리카락만 알파) 를 만든다.
• 표정 프레임은 머리와 완전히 같은 위치·크기·각도, 눈/입/눈썹만 변경(프레임 교체용).
• 민트/틸 기본색으로만 생성 — 밝음/어두움/파랑/노랑/붉은/은색은 코드에서 hairmask로 틴트.
• 투명 배경(alpha)·흰배경 금지·헤어 림라이트 금지·헤드폰/텍스트/로고 없음·목 아래 겹침 여백.

▶ 사용 안내 (당신용)

• 📋 코드블록만 붙여넣기. 먼저 short부터 만들어 파이프라인을 검증하고 나머지 모양으로 확장.
• 표정 SET은 도구 프레임 제한 시 목록을 나눠 요청. 표정은 많을수록 좋음.
• 아래는 short 기준 전체 블록. 다른 모양은 프롬프트의 헤어 묘사만 바꿔 동일하게 반복(저장 접두어 sori_head_<shape>_…).
  • short → mint/teal chin-length bob, volume bang
  • long → long straight mint/teal hair past the shoulders
  • waveS → short wavy mint/teal hair (soft curls, chin-to-neck length)
  • waveL → long wavy mint/teal hair (soft curls, past the shoulders)

---

---

§S-1. 머리 (기본 neutral) · 저장: sori_head_short.png · 첨부: sori_sheet.png

📋 붙여넣기 ↓

Using the attached reference sheet, output ONLY the HEAD of the SAME woman 이소리 (East-Asian face, brown eyes,
gentle confident smile), with mint/teal chin-length BOB hair with a volume bang. NO headphones, NO hair
accessory. Front view, head level, cleanly isolated at the neck with a small overlap margin below the chin for
rigging. Thin clean anime semi-real linework matching the sheet. Output a PNG with a FULLY TRANSPARENT
background (alpha) — no white, no shadow, no rim light / glowing outline on hair. Head only, no headphones, no
text, no logo. Avoid: text, watermark, thick outlines, white background, deformed face.

§S-2. 표정 프레임 (SET A · 핵심 6) · 저장: sori_head_short_neutral/_blink/_talk/_talk_wide/_smile/_positive.png · 첨부: sori_sheet.png + sori_head_short.png

📋 붙여넣기 ↓

Using the attached head reference, create expression frames of the SAME head (이소리, mint bob, no headphones).
Produce 6 SEPARATE images with the head in the EXACT SAME position, size, crop and angle as the attached head —
change ONLY eyes, mouth and eyebrows:
(1) neutral: eyes open, mouth closed, gentle confident smile;
(2) blink: eyes closed, mouth closed;
(3) talk: eyes open, mouth slightly open mid-speech;
(4) talk_wide: eyes open, mouth wide open (emphatic speech);
(5) smile: eyes happy, warm open smile;
(6) positive: bright excited eager look, raised cheeks (nod / yes feeling).
Keep the EXACT same face identity and framing. Thin clean anime linework. Output PNGs with a FULLY TRANSPARENT
background (alpha) — no white, no rim light on hair, no headphones. Head only, no text. Avoid: text, watermark,
inconsistent framing, white background, deformed face.

§S-3. 표정 프레임 (SET B · 리액션 6) · 저장: sori_head_short_negative/_confused/_wink/_surprised/_laugh/_thinking.png · 첨부: sori_head_short.png

📋 붙여넣기 ↓

Using the attached head reference, create MORE expression frames of the SAME head, EXACT SAME position/size/
crop/angle — change ONLY eyes, mouth, eyebrows:
(1) negative: downturned brows/mouth, gentle "no" feeling;
(2) confused: one eyebrow up, small uncertain mouth, puzzled "?" look;
(3) wink: one eye winking, playful smile;
(4) surprised: wide eyes, small open "oh!" mouth;
(5) laugh: eyes closed happy arcs, big laughing smile;
(6) thinking: eyes glancing up, lips pursed "hmm".
Keep the EXACT same face and framing. Thin clean anime linework. Output PNGs with a FULLY TRANSPARENT background
(alpha) — no white, no rim light, no headphones. Head only, no text. Avoid: text, watermark, inconsistent
framing, white background, deformed face.

§S-4. 표정 프레임 (SET C · 브랜딩 이모트 8) · 저장: sori_head_short_cool/_love/_shy/_sad/_pout/_sleepy/_proud/_playful.png · 첨부: sori_head_short.png

📋 붙여넣기 ↓

Using the attached head reference, create BRANDING EMOTE expression frames of the SAME head, EXACT SAME
position/size/crop/angle — change ONLY eyes, mouth, eyebrows (small comic accents like blush or sparkle only
where noted, no props):
(1) cool: half-lidded confident look, tiny smirk;
(2) love: heart-shaped sparkling eyes, soft happy smile;
(3) shy: light cheek blush, bashful small smile, eyes glancing away;
(4) sad: teary downturned eyes, small frown;
(5) pout: puffed cheeks, sulky pout;
(6) sleepy: droopy half-closed eyes, small yawn;
(7) proud: chin up, satisfied confident smile, closed happy eyes;
(8) playful: one eye wink + playful tongue out.
Keep the EXACT same face and framing. Thin clean anime linework. Output PNGs with a FULLY TRANSPARENT background
(alpha) — no white, no rim light, no headphones. Head only, no text. Avoid: text, watermark, inconsistent
framing, white background, deformed face.

▶ (선택) 더 추가 후보: determined(결의)·angry(발끈)·curious(호기심)·sing(노래)·starstruck(반짝)·crying_comic(폭풍눈물).

§S-5. 헤어 마스크 (색상 코드용) · 저장: sori_hairmask_short.png · 첨부: sori_head_short.png

📋 붙여넣기 ↓

Using the attached head, output a MATTE/MASK image the EXACT same size and position as the head, showing ONLY
the HAIR region filled solid white on full transparency — everything else (face, skin, neck, background) fully
transparent. This is an alpha mask of the hair only, for recoloring in code. No shading, no gradient, just a
clean solid-white hair silhouette. Output a PNG with a FULLY TRANSPARENT background (alpha). No text.

---

---

§L / §WS / §WL. 다른 머리 모양 (long · waveS · waveL)

▶ 위 §S-1 ~ §S-5 블록을 그대로 반복하되:

1. 프롬프트의 헤어 묘사 문구만 아래로 교체
   • long : long straight mint/teal hair past the shoulders
   • waveS: short wavy mint/teal hair, soft curls, chin-to-neck length
   • waveL: long wavy mint/teal hair, soft curls, past the shoulders
2. 저장 접두어를 sori_head_long_… / sori_head_waveS_… / sori_head_waveL_… (마스크는 sori_hairmask_<shape>.png)로.
3. 첨부는 해당 모양의 §-1 머리(예: sori_head_long.png)를 §-2~§-5에 첨부.

색상: 위 4모양 × (밝음·약간어두움·파랑·노랑·붉은·은색)은 이미지로 만들지 않는다. hairmask로 코드에서 틴트.