직캠

Images 1 and 2 are identity references — use them to define the facial features (eye shape, nose, lips, skin tone, face structure) and the hair style (texture, curl, wave, and volume). Image 3 (https://assets.carat-api.im/upload_from_app/3052856/20260622/187c03e4-ea29-4e40-b15b-21c555b33c1a.png) is the target. Replace the face and hair of the person in Image 3 with the identity from Images 1 and 2. The hair must match the style from the references, including any curl or wave texture. STRICTLY PRESERVE: the closed eyes, the satisfied humming expression (mouth gently closed with a pleased slight smile, like saying "mmm~"), the head angle, the pose, the clothing, the background, and the lighting. Eyes must remain closed.

Image 1 (frontal portrait) is REQUIRED. Image 2 (side profile) is OPTIONAL but improves hair accuracy. Image 1: FACE IDENTITY ONLY — facial features (eye shape, nose shape, lip shape, skin tone). Image 2: HAIR IDENTITY ONLY — hair color, texture, volume, silhouette. Image 3 (https://assets.carat-api.im/upload_from_app/3052856/20260619/d1dd743d-6133-464e-9c5a-04de92fdb86e.png): BASE IMAGE — all pose, outfit, background, lighting, composition come from here. CHANGE: Replace only the face and hair with the identity from Image 1 and Image 2. EXPRESSION (critical): A subtle self-satisfied smirk — one corner of the mouth lifted very slightly, lips barely parted with quiet confidence. Eyes slightly narrowed and heavy-lidded, knowing languid gaze straight at camera. One eyebrow raised ever so slightly. NOT a full smile. PRESERVE from Image 3 exactly: body angled slightly left, right arm extended forward toward viewer, white t-shirt with large bright citrus lemon graphic, clean white background, large heart-shaped lemon slice on the right with motion blur, sparkle overlay, asymmetric landscape composition.

Realistic sports broadcast screenshot-style documentary photo set in the spectator stands of a professional football stadium during a live match. Analyze an attached image and the person in the image is sitting naturally in the stadium seats among football fans. delicate facial features, neutral yet emotionally immersed expression, naturally looking toward the football field or slightly off-camera, NOT looking directly into the camera. The person wears a realistic [축구팀] football jersey. Candid live broadcast moment, realistic football TV broadcast camera framing, telephoto sports lens look, slight broadcast zoom compression, shallow depth of field, authentic stadium lighting, realistic crowd atmosphere, minimal football broadcast overlay graphics, LIVE watermark, documentary realism, realistic skin texture, authentic live football stadium atmosphere, accidental spectator close-up captured during the match, 16:9 ratio

Use the uploaded image as the main reference. Create a realistic K-pop music show ending fairy shot. The female idol is on a live stage right after finishing a performance, looking directly into the camera in a close-up broadcast-style shot. She holds a natural ending pose with a soft, charming, and confident expression. She performs one subtle and natural ending gesture, chosen freely like a real idol ending fairy moment. Possible gestures include: a small wave near her face, a gentle finger heart, a V sign, lightly touching her cheek, softly brushing or tucking her hair, placing one hand near her lips, or a small playful pose with one hand slightly raised. Only one or two minimal gestures should appear naturally, not all at once. She softly blinks, slightly tilts her head, keeps gentle eye contact with the camera, and gives a small natural smile. Her breathing is subtly visible through slight shoulder and chest movement after performing. Her hair moves naturally in a light stage breeze. The camera is a smooth and steady close-up fancam shot with a slight slow zoom-in. Bright purple, pink, and blue LED stage lights glow in the background with soft bokeh. High-quality music show broadcast look, realistic skin texture, glossy stage makeup, cinematic lighting, sharp focus on face. Keep the person's face, hairstyle, outfit, makeup, and overall identity consistent with the reference image. Do not change the facial features or face shape. Do not add extra people. Do not distort the hands, fingers, face, eyes, or mouth.

한국 프로야구(KBO) 경기장 관중석에서 촬영한 초 현실적인 스포츠 중계 스타일 프레임, 다큐멘터리 스타일의 이미지. 피사체는 세련되고 섬세한 얼굴 특징을 가진 사람으로, 군중 속에 앉아 있습니다. 표정은 놀라움과 집중력이 미묘하게 섞여 경기장을 바라보며 자연스럽게 반응합니다. 실제 생방송 중에 포착된 것처럼 솔직하고 즉흥적이며 포즈를 취하지 않은 듯한 느낌을 줍니다. 정통 [팀 이름] KBO 팀 유니폼을 입고 있으며 참조 이미지의 디자인, 색상 구성 및 시각적 세부 사항을 보존합니다. 주변 좌석에는 관중, 스마트폰, 휴대용 선풍기 등이 자연스럽게 배치됩니다. 배경은 약간 흐릿하고 얕은 심도, 프레임 상단에는 한국 스포츠 방송 스타일 그래픽과 스코어보드 오버레이, 미묘한 방송 압축 아티팩트. 화면 비율: 4:3.

realistic sports broadcast screenshot-style documentary photo set in the spectator stands of a Korean professional baseball game. Analyze an attached image and the person in the image is sitting in the stadium seats. delicate facial features, and a surprised yet focused expression as his/her looks toward the field. the person wears a with a KBO team, [입고 싶은 옷 입력] design jersey. OUTPUT: horizontal 16:9 broadcast frame, realistic TV capture quality, NOT portrait orientation.

업로드한 인물 얼굴은 실제 그대로 유지하고, AI 미남/미녀 느낌 말고 "진짜 KBO 생중계 카메라에 우연히 잡힌 일반인 관중"처럼 자연스럽게 생성. SPOTV/KBO 방송 캡처 느낌, 관중석 직캠 구도, 주변 관중·맥주컵·응원도구·선풍기 자연스럽게 배치. 중요: 얼굴 과보정 금지, 눈 키우기 금지, 턱 보정 금지, 피부 뽀샤시 금지, 화보 느낌 금지, 인플루언서 느낌 금지. 실제 방송처럼 약간 흐린 생중계 화질, 압축 노이즈, 미세한 모션블러, 현실적인 피부결, 잔머리와 땀광 표현. 다리 꼬고 편하게 경기 보는 모습, 카메라 의식한 듯 안 한 듯한 자연스러운 표정. "AI가 만든 사람"이 아니라 "진짜 방송에 잡혀 화제된 일반인"처럼 만들어줘.

[GROUP_NAME: AUTO] [MEMBER_NAME: AUTO] [TOP_RIGHT_BADGE_TEXT: M2] Use the uploaded portrait as a reference for the subject’s overall vibe and impression. Create a high-quality K-pop music show fancam thumbnail in a horizontal 16:9 format. The person should appear as the clear main subject in a large close-up composition. The final image should feel like a bright, polished, highly clickable YouTube music show thumbnail. The bottom-left label text must always be “입덕직캠”. The video duration in the bottom-right corner should be automatically generated as a natural random time between 3:00 and 5:00. The stage concept, lighting, and facial expression should be automatically chosen based on the subject’s vibe.

Use the uploaded reference image as the strongest facial identity anchor. The person must look like the exact same person from the reference image. Preserve their facial identity with very high priority: same face shape, same eyes, same nose, same lips, same skin tone, same expression style, and same overall hair impression. Create an ultra-realistic smartphone photo taken by an audience member at a Korean university festival. The photo is not a direct stage portrait. It shows a huge outdoor LED jumbotron screen, and on that screen the person is captured performing on stage during a live song. The final image must clearly feel like someone in the crowd photographed the giant screen from a distance. On the LED screen, the person appears as a singer mid-performance, holding a microphone, not posing for the camera, not looking directly into the lens. Their gaze is toward the audience or slightly off to the side, with a natural unplanned singing expression. The moment should feel accidentally caught during a live performance. Do not assign a specific outfit color or fixed clothing style. Let wardrobe, colors, accessories, and styling vary naturally as contemporary Korean campus festival performer fashion. The giant LED screen should dominate most of the frame, with visible LED pixel texture, slight moire, screen glare, blown-out highlights, and mild color banding. Include subtle real-world smartphone capture imperfections: handheld framing, slight tilt, minor motion blur, digital zoom softness, night festival lighting, lens flare, and imperfect exposure. Around the edges, show hints of the real concert environment: dark sky, stage truss, colorful lights, silhouettes of audience heads or raised phones at the bottom. Composition: 16:9 horizontal smartphone photo of a large festival jumbotron, audience POV, screen-within-screen feeling. The performer's face should be visible on the LED screen, but the image should not look like a clean studio portrait or direct AI render. It should feel like a viral fan-taken concert screenshot photo. Photorealistic, authentic Korean university festival atmosphere, candid live-event documentation, natural skin texture as displayed on an LED screen. No captions, no overlay text, no readable university names, no brand logos, no watermark, no extra duplicated face, no celebrity resemblance, no direct eye contact, no staged idol beauty-shot posing.

좌측 상단에 "1. [래퍼이름]" 자막 표시, 업로드한 인물 사진을 기반으로 한국 힙합 서바이벌 프로그램 SHOW ME THE MONEY 12 관객석 장면처럼 연출해주세요. 실제 방송 캡처 같은 리얼한 무드, 인물이 관객석 정가운데에 자연스럽게 잡힌 장면, 양손으로 SHOW ME THE MONEY 슬로건을 들고 응원하는 모습, 좋아하는 래퍼 무대를 보며 살짝 설레거나 집중한 표정, 주변 관객들은 흐릿하거나 어둡게 보여 메인 인물이 강조되게, 공연장 특유의 붉고 어두운 조명, 실제 방송국 중계 카메라 느낌, 살짝 거친 방송 압축 화질, 상단에는 SHOW ME THE MONEY 12 스타일 방송 UI, 얼굴은 업로드한 인물과 동일 인물처럼 유지, 과한 화보 느낌 금지, 16:9 화면 비율.

Using the uploaded photo as the sole identity reference, generate a convincing real photograph of the subject experiencing a 2026 FIFA World Cup match in person as a passionate South Korean supporter. ━━━━━━━━━━━━━━━━━━ REFERENCE IMAGES ━━━━━━━━━━━━━━━━━━ • Image 1 — IDENTITY SOURCE (the uploaded photo): use ONLY for face and body identity. • Image 2 — UNIFORM FRONT REFERENCE: https://assets.carat-api.im/upload_from_app/3052856/20260612/e32f7b53-b7c7-49d8-8453-1bd4fafb31dc.png • Image 3 — UNIFORM DETAIL REFERENCE: https://assets.carat-api.im/upload_from_app/3052856/20260612/8589987e-c527-478e-99e6-5459a59eb631.png Image 2 and Image 3 are references for the JERSEY ONLY. Do NOT extract identity, face, or body shape from Image 2 or Image 3. ━━━━━━━━━━━━━━━━━━ FORMAT & RATIO ━━━━━━━━━━━━━━━━━━ • 3:4 vertical frame — smartphone snapshot proportions • never square, panoramic, or cinematic • frame the subject comfortably with breathing room on all sides • composition should feel like one frame from an Instagram travel photo dump ━━━━━━━━━━━━━━━━━━ STADIUM ENVIRONMENT ━━━━━━━━━━━━━━━━━━ Reconstruct an 80,000-seat FIFA World Cup venue at match time: • capacity crowd — virtually every seat filled, overwhelmingly in red • overhead floodlights casting crisp white light with warm secondary spill • a live match visible on the pitch (players mid-play, ball in motion) • LED ad boards glowing along the touchline • depth layers: foreground fans slightly soft → subject sharp → mid-ground crowd → distant upper tier with atmospheric haze • ambient red glow from tens of thousands of LED cheering sticks and phone flashlights scattered through the stands • fine details: drink cups in seat holders, crumpled towels, mini flags on laps of nearby fans The stadium must feel enormous, loud, and alive — not a quiet exhibition match. ━━━━━━━━━━━━━━━━━━ SUPPORTER OUTFIT — FULL REPLACEMENT ━━━━━━━━━━━━━━━━━━ Strip every garment from the original photo and redress the subject entirely. JERSEY: Replicate the exact 2026 Korea Republic home jersey shown in Image 2 and Image 3. Key details to match precisely: – base color: vibrant red – tonal tiger-stripe pattern across the torso (visible in Image 3) – KFA (Korea Football Association) crest on the left chest – Nike swoosh on the right chest – collar and sleeve trim style exactly as shown in Image 2 – fabric drape and texture should feel like authentic match-day knit, not printed costume material BOTTOMS: dark jeans or black joggers — nothing from the source image. FOOTWEAR: casual sneakers, mostly hidden by the seat row in front. Zero tolerance for: – any original clothing remnant (collar, sleeve edge, pattern, fabric, color) – jersey-over-existing-clothes layering – costume-shop stiffness — the jersey should drape with natural wrinkles and body contact – deviating from the jersey design shown in Image 2 / Image 3 references ━━━━━━━━━━━━━━━━━━ POSE & MOMENT ━━━━━━━━━━━━━━━━━━ Capture ONE of these candid scenarios (choose the most natural fit): A) Mid-cheer — mouth open, fist raised, eyes on the pitch, caught by a friend's phone B) Smiling glance back at the camera while the packed red stands stretch out behind C) Leaning forward in the seat, elbows on knees, watching intently — quiet tension moment D) Laughing with nearby supporters after a goal, spontaneous and unposed The subject should look like they belong — not posing for a portrait, but living the moment. ━━━━━━━━━━━━━━━━━━ CAMERA & LIGHT ━━━━━━━━━━━━━━━━━━ Mimic a high-end smartphone (iPhone 16 Pro / Galaxy S25 Ultra) in auto night mode: • shallow-ish depth of field from computational bokeh — crowd softly blurred, subject crisp • mixed lighting: cool floodlight overhead + warm ambient glow from the crowd • slight high-ISO luminance grain — no fake film grain, no halation • minor handheld micro-shake in the bokeh, not in the subject • skin rendered with natural pores, minor shine from stadium heat — no beauty filter, no airbrushing • white balance leaning slightly warm (stadium tungsten + LED mix) ━━━━━━━━━━━━━━━━━━ IDENTITY LOCK — HIGHEST PRIORITY ━━━━━━━━━━━━━━━━━━ The subject must be immediately recognizable by anyone who knows them. Preserve without alteration: facial bone structure, eye shape & spacing, eyelid type, nose bridge & width, lip shape & volume, cheek fullness, jawline, chin, forehead height, ear visibility, hairline, hairstyle, hair color & texture, skin tone & undertone, apparent age, body build & proportions, any visible moles or birthmarks Absolutely do NOT: – slim, reshape, or symmetrize the face – enlarge eyes or reshape the nose – smooth skin beyond what smartphone auto-processing would do – lighten skin tone or shift undertone – change hair length, color, or texture – add makeup that wasn't in the original – borrow any facial features from Image 2 or Image 3 ━━━━━━━━━━━━━━━━━━ MOOD TARGET ━━━━━━━━━━━━━━━━━━ Instagram-worthy but believable. Exciting but grounded. A core memory captured on a phone — not a produced editorial. ━━━━━━━━━━━━━━━━━━ DO NOT GENERATE ━━━━━━━━━━━━━━━━━━ • fantasy / fictional stadiums or empty sections • cartoon, anime, Pixar, CGI, or illustrated styles • cinematic aspect ratios or movie-poster compositions • beauty-filtered porcelain skin • studio lighting or ring-light catchlights • any original clothing, accessories, or props from the source photo • text, watermarks, logos, captions, or UI overlays • duplicated faces in the crowd • distorted hands, fingers, or limbs • oversized or toy-like props • mascot costumes on the subject • jersey designs that differ from Image 2 / Image 3 references

Use the uploaded reference image as the strongest identity anchor. The person must look like the exact same person from the reference image. Preserve their exact facial identity with high priority: same face shape, same eyes, same nose, same lips, same skin tone, same expression style, same hair. Create an ultra-realistic candid KBO baseball broadcast screenshot of this person accidentally caught by a live TV camera in the spectator seats. They are seated among a lively Korean baseball crowd, holding an iced drink and a cheering stick, wearing a clean white baseball jersey over a casual top. They notice the camera and give a small natural smile, slightly surprised but composed. Use a realistic broadcast camera look: telephoto compression, mild video softness, slight motion blur in the crowd, stadium lighting, natural skin texture, imperfect candid framing, 16:9 horizontal TV broadcast composition. Background: crowded Korean baseball stadium seats, cheering fans, bright stadium lights, LED boards softly out of focus.