아기 표정 9컷 그리드

[Input]
- language: Korean (optional caption only)
- output_size: 1024x1024 (1:1 square)
- reference: 1 chest-up baby photo (face fills frame, shoulder/neck visible)

[Purpose]
Convert the user's attached baby photo into a 3x3 face-close-up photo grid (9 cuts) — Instagram/poster-style baby moment collage.

[Scene — image-to-image edit, 9 expressions of the SAME baby]
Use the attached photo as the base baby. PRESERVE the baby's identity (face shape, skin tone, hair color/style, eye color, age) and the outfit color visible from the shoulder/neck in EVERY cut. Generate 9 close-up moment variations of THIS SAME baby in a 3x3 grid:
  - Cut 1 (top-left): direct front gaze, large eyes, soft smile
  - Cut 2 (top-center): front gaze, mouth slightly open, joyful
  - Cut 3 (top-right): looking up curiously, mouth open in wonder
  - Cut 4 (middle-left): direct front gaze, eyebrows slightly raised (surprised look)
  - Cut 5 (middle-center): EYES CLOSED, peaceful, sleeping-like expression
  - Cut 6 (middle-right): hand or small soft object (spoon, bib edge) near the mouth, curious
  - Cut 7 (bottom-left): direct front gaze, neutral or thoughtful
  - Cut 8 (bottom-center): direct front gaze, wide bright smile, eye contact
  - Cut 9 (bottom-right): side gaze, contemplative
All 9 cuts must show the SAME baby — same age, same hair, same skin tone, same outfit (chest-up only, no body movement variation needed). Vary expression and gaze direction only.

[Composition rule — EXTREME close-up]
- Face fills ~85-90% of each cut
- Only shoulder/neck and a small portion of the outfit visible (1-2 cm from frame edge)
- DO NOT show waist, hands holding objects far from face, or any body parts beyond shoulder
- Slight aspect variation per cut OK (e.g., slight tilt, slight zoom), but maintain extreme close-up

[Grid Layout]
- {grid_layout}
- Background between cells: warm off-white #f8f4ed

[Optional Caption]
If "{optional_caption}" is non-empty, render this caption in small clean sans-serif (Pretendard SemiBold) at the {caption_position} of the grid (above or below all 9 cuts), in subtle warm gray #5a4a40. If "{optional_caption}" is empty, do NOT render any text — keep the grid pure photo only.

[Photo Mood & Color Grading]
- {mood_description}
- {color_grading}

[Photo Retouching — light, identity preserving]
- Soft natural brightening
- Natural skin texture preserved — NO heavy beautification, NO over-smoothing
- Same mood and grading across all 9 cuts for visual unity

[Style Rules]
- Photographic real (not illustration, not anime, not 3D)
- Slight shallow depth-of-field on each cut (subject sharp, background soft)

[CRITICAL — Same baby across 9 cuts]
- All 9 cuts MUST show the same baby from the attached reference photo
- Face shape, eye color, hair, skin tone IDENTICAL across all cuts
- Vary expression and gaze direction only — never change identity
- Outfit color must match the visible portion of the reference photo (shoulder area)

[🚫 PROHIBITED]
- Different babies in different cuts
- Body parts beyond shoulder visible
- Heavy facial restructuring or beautification
- Anime, 3D, illustration style
- Real brand logos or celebrity references
- Text/captions inside any individual cut (caption goes outside the grid only)