聊天GPT
日落的屋顶自信畫像"
為聊天GPT 編寫。不同型號的結果和參考照片支援可能有所不同。
使用提示詞前
列出的工具:聊天GPT。完整說明:3,034 個字元。 1 個提供的範例影像。
This prompt mentions reference images. Read its instructions for the required number of photos and identity constraints. We have not independently tested this prompt. Tool labels and supplied images are not a guarantee of compatibility or output quality.
完整提示詞
CRITICAL RULE #1 (read first, applies above all else): The face in the
output MUST be the exact same person as the uploaded reference photo.
Do not generate a new face, do not beautify, do not idealize, do not
blend with a generic face. If you are unsure whether the face matches,
default to making it MORE identical to the reference, not less.
IDENTITY LOCK (exact match required): Match the reference photo exactly
for: face shape and bone structure, eye shape/size/color/spacing,
eyebrow shape, nose shape and width, lip shape and fullness, chin and
jawline shape, cheekbone structure, skin tone and undertone, any
visible marks, moles, or asymmetries. These are not suggestions — they
are hard constraints. Zero deviation allowed.
DO NOT COPY FROM REFERENCE (only these things should change): expression,
eye gaze direction, head angle/tilt, and body pose should NOT match the
reference photo — generate these freshly based on the scene below. To
be clear: FACE IDENTITY = locked exactly. EXPRESSION/POSE/ANGLE = freely
generated and different from the reference.
EMOTIONAL INTENT (design goal): An effortless, confident golden-hour
rooftop moment — the kind of photo she'd genuinely want to post,
looking like her best real self, not a different or idealized person.
LIGHTING SETUP: Key light — warm golden-hour sun from camera-left, low
angle. Fill — soft ambient sky light balancing at roughly 2:1. Rim
light — sun grazing her hair from behind. Catchlights warm and soft.
LENS & OPTICS: Simulate an 85mm lens at f/2.0, flattering facial
compression, soft creamy bokeh background of a city skyline, gentle
vignette.
FILM/COLOR SCIENCE: Warm golden-hour film response, soft highlight
rolloff, gently lifted shadows, natural true-to-reference skin tone —
color grading must not shift her actual skin tone from the reference.
Style: A flattering, wearable outfit (flowy dress or co-ord set in a
warm tone), minimal natural makeup, hair softly styled — realistic and
relatable, not an exaggerated glam transformation.
NEW POSE & ANGLE (must differ from reference, face identity unchanged):
Standing near the rooftop ledge, one hand lightly touching her hair as
it lifts in the breeze, body angled at three-quarters to the camera,
head turned back over one shoulder toward the lens.
NEW EXPRESSION (must differ from reference, face identity unchanged):
A soft, warm, genuine smile with eyes gently crinkled — natural joy,
not a stiff copied expression.
RETOUCH PIPELINE: Smooth tonal transitions while fully preserving fine
pore-level texture and the reference face's exact proportions. No
flattening, no waxy skin, no facial reshaping of any kind.
FINAL VERIFICATION (mandatory self-check before output): Compare the
generated face side-by-side against the reference photo's face shape,
eyes, nose, lips, and jawline. If any facial feature does not match
the reference person, discard and regenerate the face to match exactly
— only the expression, pose, and angle are allowed to differ.複製完整的提示並根據要求使用您自己的參考照片。
此提示提供的範例影像。此提示尚未經過我們網站的測試。


