Phimetra does not retouch photos. It measures. Each diagnostic is based on anatomical references validated since 1957, applied by an AI engine trained on thousands of faces.
The engine detects anatomical landmarks of the face: Ricketts points, midlines, mandibular contours. Each point is positioned with sub-pixel precision.
Ratios are computed according to the Marquardt protocol: golden ratio (1.618), facial symmetry, facial thirds, zygomatic projection. Seven independent metrics, all explainable.
Each facial zone is assessed individually. The diagnosis identifies the zones where a treatment would have the most impact: lips, cheekbones, dark circles, jawline, forehead. No global suggestion — zone by zone.
The simulation generates a before/after image on the zones selected by the patient. The engine respects the bone structure (ControlNet) to avoid unrealistic deformations. This is not a filter — it is a projection based on injectable volumes.
Photo not retained
The image is processed in RAM then immediately deleted. No storage, no third-party cloud.
GDPR compliant
Explicit consent, audit log of acceptances, right to erasure. Data hosted in France.
Full transparency
Each measurement is visible in Glass Box mode. The practitioner sees landmarks, ratios and the calculation method.
No false promises
The simulation is indicative. Only a qualified practitioner can issue a medical diagnosis and propose a procedure.
Simulations are indicative and do not constitute medical advice. Only a qualified practitioner may issue a diagnosis and recommend a procedure.