AI Face Reading With Samudrika and Mian Xiang

What Is Face Reading?

Physiognomy reads the structure of the face as a map of disposition. Mian Xiang (面相) divides the face into three zones — the forehead for thought and early formation, the middle face for effort and the working years, the lower face for endurance and later life — and locates twelve palaces on it: the career palace at the centre of the forehead, the wealth palace on the nose, the siblings palace on the brows, the spouse palace at the temples, the servants palace on the chin and jaw, among others. The classical sources are the Ma Yi Shen Xiang (麻衣神相, Song dynasty) and the Liu Zhuang Xiang Fa (柳庄相法, Ming dynasty). The Indian Samudrika Shastra reads the same features — the lalāṭa (forehead), netra (eyes), nāsikā (nose), oṣṭha (lips) and chin — as marks of temperament and fortune. ProxaAI uses both, and says which tradition each indication comes from.

What ProxaAI Computes From Your Face

The pipeline is the same three-step contract as the palm reading, with the measurements done in code:

• Guided capture. An oval on the screen shows where to place your face; the page tracks 478 face landmarks live, checks that you face the camera squarely (yaw and pitch), and refuses a blurred, dark or badly lit frame before uploading.
• Measured geometry. The three zones, the face's breadth and the proportions the landmarks measure exactly — the spacing of the eyes, the width of the mouth, the philtrum — are computed in the photo's own pixels and judged against MediaPipe's average face, not estimated by a model: a zone is dominant only when it is long for its own share of the face. A left–right symmetry index is not read — a slight turn of the head moves it more than faces differ.
• Feature extraction, once. The first question after uploading sends the image to a vision model that fills a fixed schema — face shape, forehead height and width, brow shape and density, eye shape, nose bridge and tip, lips, chin, ears (judged against the brow and nose lines), cheekbones — with a confidence for each, and each must clear its own confidence bar; hair over the forehead or glasses over the eyes simply make those regions unreadable.
• Rule-based reading. Some fifty cited rules — zones, face shapes, palaces, and Samudrika combinations such as the thinker's, leader's and steward's profiles — fire only on the extracted features, and twelve Mian Xiang age-position scales, one per region of the face and one per zone, attach a span of years to a position-fortune reading that fired on that region — the fortune of the years the region governs, judged from its form — and never to a trait of character. The face reader cites only fired rules and fired windows; a validator strips anything else.

How the Reading Uses Your Face

When face evidence exists and your question touches character, career, relationships or money, the turn planner adds the face reader to the voices it consults. It runs beside the Vedic astrologer, the numerologist and, if you have captured your palms, the palmist. The synthesis weighs each voice by the domain of the question — face reading carries weight on character and career, none on longevity or health (it may only name an area that asks for extra care, as a forewarning) — and reports where the readings agree. When the face does place something in time, it is an approximate age window from the Mian Xiang positions, set beside the dasha timing as a corroborating voice, never in place of it. The methodology page describes the weighting.

What the Face Reader Will and Will Not Say

It reads tendencies: a high forehead as the career palace's capacity for study and public responsibility, prominent cheekbones as will and authority, a full nose tip as a generous relation to resources, wide-set eyes as tolerance in partnership. It does not estimate your age from the photo, and will not comment on your gender, ethnicity, skin, health, mood, weight or attractiveness — the extraction prompt forbids describing them and the schema has no field for them. What it may place in time is an age window from the classical positions: Mian Xiang attaches a span of years to each region of the face — the ears to childhood, the forehead to youth, the brows and eyes to the thirties, the nose to the forties, the mouth to the fifties, the chin to later life — and where a position-fortune reading of that region fired (the form of the region: a high, full forehead, a straight bridge, a firm chin), the face reader gives that span as an approximate age window, names the chart it comes from and says whether it lies behind you, is open now or lies ahead of you. Never a date, month, year or festival, never lifespan, death or a diagnosis, and never an illness or an accident placed in time; it may name an area that asks for extra care of health or against accidents, as a forewarning. The philtrum and every reading that touches the family are read for character alone — never for health, never for the timing of a birth — and the reading says so when it uses them.

Timing From the Face: The Age Positions, Read as Windows

Mian Xiang maps a life onto the face: the hundred age positions of the Ma Yi Shen Xiang chart run from the ears in childhood (ages 1 to 14), across the forehead (15 to 30), the brows (31 to 34), the eyes (35 to 40), the nose (41 to 50) and the mouth and philtrum (51 to 60), to the chin and jaw in later life (61 onward), and the three zones (San Ting) stand for youth, the middle years and age. The exact point names and single-year assignments vary by edition, which is why ProxaAI encodes the chart only at the level of a region and widens any band narrower than seven years. It uses this the way it uses every rule — only when the feature is readable and the rule that carries the position fired — and with two honest limits. First, a photograph supports the region, not the point: the face reader gives the span the chart attaches to a region as an approximate age window, never a single year picked from the hundred. Second, it is told your age bracket in words and your life context, never your date of birth or age; whether a span is behind you, open now or ahead of you is worked out in code from your age and handed to it. A span of ten years or more is stated only whole ("roughly between 41 and 50"), never as a single year inside it, and it cannot be turned into a date, a month, a year or a festival; a validator removes any sentence that tries. Colour and complexion, the chart's first sign of a position's fortune, are never read from a photograph, so a window rests on the form of a region alone. What it never places in time: lifespan, death, illness, accidents, a birth or any event the ethics rules forbid. It may name an area that asks for extra care of health, as a forewarning, never a diagnosis and never with an age. Nothing here estimates how old you look; the windows come from the chart, not from the photo.

Privacy and Consent

Face capture is optional, adults only, and switched on only after you give explicit consent on the capture page. Photographs are stored in a private, region-locked bucket and deleted automatically after 30 days; the extracted features stay with your profile until you remove them from the dashboard. Nothing is used for training, identification or matching, and no one at ProxaAI can view your images. The privacy policy and terms of use set this out in full.