The hardest thing in AI content — two consistent faces in a single candid frame, each held to its own reference.
The result
Works in
Nano Banana / any image model that takes two face references
How to use
Attach both references. Paste the prompt. The model composes both into one candid cafe frame.
Swap this
Change the reference face to your own character and keep the structure — that's the whole trick.
Steal My Prompt
Two women at a cafe table. Left: a blonde woman, long wavy hair, green-grey eyes, fair skin, narrow nose bridge, full lips. Right: a woman with dark straight hair, olive-tan skin, light green eyes, thick dark brows, high cheekbones, a defined jawline. Keep each face consistent to its reference: eye shape, eye set, nose shape, lips, cheekbones, face oval, age. Do not average, do not blend, do not "improve", do not de-age.
FRAME: a tight chest-up two-shot, both heads large and filling the upper half of the frame. The table is visible only at the lower edge — a cappuccino cup, the edge of a plate with a croissant, crumbs, blurred in the foreground. Camera at eye level, close, as if shot across the table from arm's length.
POSE: the dark-haired woman leans slightly toward the other, telling her something, hand raised mid-gesture. The blonde holds the cup with both hands at her chest, laughing, looking at her friend. Neither looks at the camera.
WARDROBE: blonde — a cream cable-knit polo-collar top with short sleeves, gold rings, a gold watch. Dark-haired — a blue denim shirt with rolled sleeves, a thin gold chain, a silver bracelet.
SCENE: a bright cafe, a large window on the left, a stone wall and plants blurred in the background, soft daylight.
PHOTOREALISM: shot on an iPhone, light grain, honest light, skin with pores, vellus hairs, stray strands, natural sebum shine. No retouching, no filters, no skin smoothing, no studio light.
Emotion, intensity 3: the blonde — a crinkle-eyed squint, mouth open in a laugh, inner brow corners raised; the dark-haired woman — one brow higher, lips mid-sentence, lively expression.
No facial or hand deformation, no extra fingers, no face swapping, no averaged "model" faces, no text or watermarks.
The whole stack, one link → syntx.ai
AI-generated & conceptual. Product names shown for creative reference only; not affiliated with or endorsed by any brand. Characters are original synthetic creations of Victory AI Creative Studio.