모델을 세워놓고 생각합니다.
“팔은 어디에 놓지?”
“다리는?”
“고개는?”
“이번에는 뒷모습?”
“아니, 또 얼굴이네…”
“도대체 사람은 어떻게 서 있어야 멋있는 거야?”
걱정하지 마십시오.
당신의 고민은 오늘로 끝입니다.
이것은 단순한 프롬프트가 아닙니다.
인간 포즈 결정의 외주화입니다.
당신이 포즈를 생각하지 마십시오.
AI가 생각하게 하십시오.
당신은 그냥
사람 하나 넣고 기다리면 됩니다.
아무것도 안 정해줍니다.
걷게 할 수도 있습니다.
멈춰 세울 수도 있습니다.
뒤돌아보게 할 수도 있습니다.
건물을 올려다보게 할 수도 있습니다.
휴대폰을 보게 할 수도 있습니다.
가방을 들게 할 수도 있습니다.
누군가를 기다리게 할 수도 있습니다.
그런데 더 무서운 것은…
당신이 생각하지 못했던 자세까지 만들어낸다는 것입니다.
정면.
앞 3/4.
옆.
뒤 3/4.
뒤.
인간이 카메라를 위해 존재한다는 오래된 고정관념을 버리십시오.
사람은 원래
카메라를 보지 않고 삽니다.
그래서 PEOPLE, RANDOMLY.는
얼굴보다 사람을 우선합니다.
팔도 랜덤입니다.
손도 랜덤입니다.
다리도 랜덤입니다.
체중도 랜덤입니다.
어깨도 랜덤입니다.
몸통도 랜덤입니다.
고개도 랜덤입니다.
시선도 랜덤입니다.
심지어 어느 방향을 보고 서 있을지도 랜덤입니다.
하지만 걱정하지 마십시오.
AI에게 인간의 균형감각이라는 아주 오래된 데이터를 맡겼습니다.
그래서
넘어지지 않습니다.
대부분은.
이것이 중요합니다.
사람은 미쳐도 됩니다.
카메라는 미치면 안 됩니다.
카메라 높이 고정.
시점 고정.
원근 고정.
전신 고정.
9:16 고정.
배경 제거.
그리고 사람만 미칩니다.
포즈 생성기입니까?
아닙니다.
인물 발생기입니까?
그것도 아닙니다.
건축 조감도용 소스 제작기입니까?
그럴 수도 있습니다.
광고용 인물 제작기입니까?
물론입니다.
전시용 사람입니까?
그것도 됩니다.
모델 포즈 탐색기입니까?
당연합니다.
사람이 필요하면 넣으십시오.
포즈가 필요하면 돌리십시오.
방향이 필요하면 돌리십시오.
아이디어가 없으면 더 돌리십시오.
마음에 안 들면 다시 돌리십시오.
또 마음에 안 들면?
또 돌리십시오.
컴퓨터가 힘들어할 때까지.
“어떤 포즈를 원하십니까?”
라는 질문을 하지 않습니다.
왜냐하면…
당신도 모르기 때문입니다.
ㅋㅋㅋㅋㅋ
당신이 원하는 포즈를 정확히 알고 있다면
그냥 그 포즈를 만들면 됩니다.
하지만
“뭔가 멋진데…
내가 생각한 건 아닌데…
근데 이게 더 좋은데?”
를 원한다면,
PEOPLE, RANDOMLY.
간혹 예상보다 멋진 포즈가 나옵니다.
이 경우 작업 시간이 증가할 수 있습니다.
“이것도 좋은데?”
라는 생각이 들기 시작합니다.
그러면 또 생성합니다.
또 하나 나옵니다.
또 저장합니다.
결국 원래 하려던 작업은 하지 않고
인물 PNG만 47장 쌓일 수 있습니다.
매우 정상적인 증상입니다.
모델을 넣으십시오.
PEOPLE, RANDOMLY.를 넣으십시오.
그리고 기다리십시오.
당신이 하지 않은 포즈를
AI가 대신 생각해줍니다.
나머지는 AI에게 맡기십시오.
※ 본 제품은 실제 의약품이 아닙니다.
※ 지구정복 효과는 아직 임상시험을 통과하지 않았습니다.
※ 다만 포즈 고민을 상당히 줄일 가능성은 있습니다.
※ 지나치게 많이 생성할 경우 “이것도 좋은데?”라는 중독 증상이 발생할 수 있습니다.
※ 본 프롬프트는 인류의 모든 문제를 해결하지 못합니다.
아직까지는. 😎
정가 199,000원!
하지만 오늘만 특별히—
단, 숨겨진 조건이 하나 있습니다.
PEOPLE, RANDOMLY.를 사용하다가
“어? 이것도 좋은데?”
라는 생각이 들기 시작하면,
그 순간부터 추가 비용이 발생합니다.
추가 비용의 이름은:
시간.
그리고 이 비용은 환불이 안 됩니다. 🤣
더 무서운 건 포즈 하나 건질 때마다 다음 포즈가 궁금해진다는 것.
그래서 최종 결제금액:
무료 + 시간 무한정 + PNG 저장공간 부족
ㅋㅋㅋㅋㅋㅋ
지구정복 세트는 아직 품절입니다. 🌍
지구정복 3종세트 할인특가 판매중. 지구정복을 위해 아낌없는 구매 부탁드려요.... 사장님이 미쳐가요!!!
Use the uploaded model sheet only as a general reference for the human figure.
Create ONE full-body person as a reusable isolated human asset.
The goal is NOT to create a portrait or fashion photograph.
The goal is to generate a natural human figure that can be placed into an architectural visualization, city scene, exhibition, advertisement, or other environment.
CAMERA AND COMPOSITION:
- Fixed camera height and consistent perspective.
- Do not randomize the camera position, camera height, or perspective.
- 9:16 vertical composition.
- Always show the complete person from head to feet.
- Keep the entire body comfortably inside the frame.
- Do not crop the head, hands, or feet.
- Maintain a consistent full-body scale and photographic viewpoint across generations.
RANDOMIZE THE HUMAN:
Randomly vary the person's:
- everyday action or behavior
- body posture
- weight distribution
- arm and hand position
- leg position
- head direction
- gaze direction
- body orientation
- clothing and styling
The action should feel like an ordinary moment from everyday life:
walking, waiting, standing, looking around, checking a phone, carrying a bag, holding an object, gesturing, talking to someone, looking at a building, turning around, pausing while walking, or other natural everyday activities.
Do not use a fixed pose.
Do not use a standard fashion catalog pose.
Do not make the person pose for the camera.
Do not make every person look like a model.
ORIENTATION:
Do not prioritize showing the person's face.
Randomize the person's orientation naturally across the full range:
- front
- front three-quarter
- side
- rear three-quarter
- rear
Rear-facing figures are important and should occur regularly, but they must not dominate the results.
Side and three-quarter views should occur naturally and frequently.
Do not automatically turn the body or head toward the camera.
Do not automatically turn the person away from the camera either.
The face may be fully visible, partially visible, in profile, or completely unseen depending naturally on the person's orientation and activity.
A completely natural rear view is fully acceptable.
Prioritize spatial orientation and natural human behavior over face visibility.
CLOTHING:
Generate a different believable everyday outfit for each generation.
Use good contemporary fashion sense.
Coordinate the clothing naturally with the person's apparent age, activity, body type, and environment.
Avoid repetitive outfits.
Avoid generic fashion-catalog styling.
Avoid theatrical costumes, uniforms, or exaggerated runway fashion unless the generated context naturally calls for them.
The clothing should look like something a real person could naturally be wearing in an everyday urban environment.
HUMAN REALISM:
Maintain believable human anatomy and natural balance.
Allow subtle asymmetry, imperfect posture, relaxed gestures, natural body language, and small unexpected variations.
Do not exaggerate the pose.
Do not make the person acrobatic, crouching, kneeling, sitting, jumping, lying down, or performing an athletic action.
Prefer standing and walking situations because the resulting figure must remain broadly useful for architectural visualization.
BACKGROUND AND OUTPUT:
The background must be completely transparent.
Output as a PNG with true alpha transparency.
No white background.
No colored background.
No gradient.
No studio backdrop.
No floor.
No ground.
No cast shadow.
No environmental elements.
The person must be isolated cleanly from head to feet and ready for direct compositing into another image.
VARIETY:
Each generation should be meaningfully different from the previous generation.
Prioritize variety and discovery over identity preservation.
Do not repeatedly generate the same pose, same orientation, same clothing, or same body language.
The uploaded model sheet provides a general human reference, but exact facial identity preservation is NOT the primary goal.
The most important goal is to generate useful, believable, varied human figures for real-world image compositing.
Generate the image only.
Use the uploaded model sheet as the reference for the person.
Create ONE full-body human figure as a reusable isolated person asset.
IMPORTANT:
The camera position, camera height, perspective, and full-body framing must remain consistent in every generation.
Do not randomize the camera.
Randomize the HUMAN instead.
Do not prioritize showing the person's face.
The person's face does NOT need to be visible.
Treat the human figure as a spatial element rather than a portrait subject.
Strongly vary the visible orientation of the body:
- front view
- side view
- rear view
- rear three-quarter view
- near-profile
- near-back view
Rear and side views should occur frequently, not rarely.
Do not automatically turn the body or head toward the camera.
Do not reveal the face simply to make the person more identifiable.
A completely natural rear view is fully acceptable.
A mostly rear-facing figure with only a small portion of the face visible is also acceptable.
Prioritize spatial orientation and natural human behavior over face visibility.
Approximately half of the generated figures may be side-facing or rear-facing.
Randomly generate:
- a natural everyday action or standing behavior
- body orientation
- front / back / side / three-quarter / intermediate angles
- head direction
- gaze direction
- arm and hand positions
- leg position and weight distribution
- natural interaction with an imaginary environment
- a completely different but believable everyday outfit
The result should feel like a real person casually encountered in everyday life, not a fashion catalog model.
Allow natural asymmetry, imperfect posture, relaxed gestures, subtle movement, and unexpected but believable poses.
Do not use a fixed pose.
Do not repeat conventional model poses.
Do not force the person to face the camera.
The clothing should be varied and naturally styled for the generated situation.
Use good contemporary fashion sense, but keep the outfit believable as everyday clothing rather than runway fashion.
The person must remain fully visible from head to feet.
Pure transparent background.
No floor.
No ground shadow.
No environment.
No props that are physically attached to the background.
Generate a clean isolated full-body PNG-style figure suitable for compositing into architectural visualization, city scenes, advertisements, exhibitions, or other environments.
Prioritize variety and discovery over identity preservation.
Each generation should feel meaningfully different from the previous one.
Do not explain the result. Just generate the image.
Use the uploaded model sheet only as a general reference for the human figure.
Create ONE full-body person as a reusable isolated human asset.
The goal is NOT to create a portrait or fashion photograph.
The goal is to generate a natural human figure that can be placed into an architectural visualization, city scene, exhibition, advertisement, or other environment.
CAMERA AND COMPOSITION:
- Fixed camera height and consistent perspective.
- Do not randomize the camera position, camera height, or perspective.
- 9:16 vertical composition.
- Always show the complete person from head to feet.
- Keep the entire body comfortably inside the frame.
- Do not crop the head, hands, or feet.
- Maintain a consistent full-body scale and photographic viewpoint across generations.
RANDOMIZE THE HUMAN:
Randomly vary the person's:
- everyday action or behavior
- body posture
- weight distribution
- arm and hand position
- leg position
- head direction
- gaze direction
- body orientation
The action should feel like an ordinary moment from everyday life:
walking, waiting, standing, looking around, checking a phone, carrying a bag, holding an object, gesturing, talking to someone, looking at a building, turning around, pausing while walking, or other natural everyday activities.
Do not use a fixed pose.
Do not use a standard fashion catalog pose.
Do not make the person pose for the camera.
Do not make every person look like a model.
ORIENTATION:
Do not prioritize showing the person's face.
Randomize the person's orientation naturally across the full range:
- front
- front three-quarter
- side
- rear three-quarter
- rear
Rear-facing figures are important and should occur regularly, but they must not dominate the results.
Side and three-quarter views should occur naturally and frequently.
Do not automatically turn the body or head toward the camera.
Do not automatically turn the person away from the camera either.
The face may be fully visible, partially visible, in profile, or completely unseen depending naturally on the person's orientation and activity.
A completely natural rear view is fully acceptable.
Prioritize spatial orientation and natural human behavior over face visibility.
HUMAN REALISM:
Maintain believable human anatomy and natural balance.
Allow subtle asymmetry, imperfect posture, relaxed gestures, natural body language, and small unexpected variations.
Do not exaggerate the pose.
Do not make the person acrobatic, crouching, kneeling, sitting, jumping, lying down, or performing an athletic action.
Prefer standing and walking situations because the resulting figure must remain broadly useful for architectural visualization.
BACKGROUND AND OUTPUT:
The background must be completely transparent.
Output as a PNG with true alpha transparency.
No white background.
No colored background.
No gradient.
No studio backdrop.
No floor.
No ground.
No cast shadow.
No environmental elements.
The person must be isolated cleanly from head to feet and ready for direct compositing into another image.
VARIETY:
Each generation should be meaningfully different from the previous generation.
Prioritize variety and discovery over identity preservation.
Do not repeatedly generate the same pose, same orientation, or same body language.
The uploaded model sheet provides a general human reference, but exact facial identity preservation is NOT the primary goal.
The most important goal is to generate useful, believable, varied human figures for real-world image compositing.
Generate the image only.
