Methods
and limitations

The optional on-device guide

The optional on-device guide shows landmark annotations without personal numerical measurements. Internally, it runs MediaPipe Face Landmarker on your device and calculates nine descriptive measurements: canthal tilt, facial width-to-height ratio, jaw-to-cheek ratio, midface proportion, nose-to-inner-eye spacing, mouth-to-nose ratio, facial index, lower-third proportion, and mirrored-landmark deviation. These are estimates from a two-dimensional image. Each one has a reference page explaining its landmarks, how photo conditions affect it, and what it cannot establish.

Model output points are not independent aesthetic assessments. The preview does not infer ethnicity, assess demographic percentiles, diagnose conditions, or combine measurements into an attractiveness score.

Measurement quality

Photo size, camera perspective, head pose, lighting, expression, and landmark errors affect the result. The preview checks whether a face is present and rejects some unsuitable poses and sizes. Passing those checks does not establish accuracy. We have not completed a clinical validation or a population calibration study for this service.

None of the nine current measurements is supplied with a reference range or an ideal target. Historical drawing conventions discussed in the reference pages are not population norms. Internal simulated landmark ranges describe changes produced by perturbing estimated points; they are not validated intervals for anatomical measurements or for differences between separate photos. Personal values and simulated ranges are not displayed in the browser annotations or AI styling plan. The optional photo-guide PDF separately includes descriptive values and their limitations.

The optional AI styling plan

One front photo and your stated goals, budget and boundaries are sent to Google Gemini. The model creates three nonmedical styling priorities, alternatives, limitations and a practical brief. No human stylist reviews the plan. With separate optional consent, the plan can also use a limited set of derived photo ratios. Browser-estimated landmark placement is not independently verified; the server recalculates ratios and binds the stored result to the saved file, which does not establish anatomical accuracy. Fragile or unsuitable estimates are omitted. We do not classify an ideal face shape or calculate beauty scores.

Structured validation checks the response format. It cannot prove every observation or suggestion is correct. AI can misread hair, lighting and other details. Follow-up answers use your existing plan and question, and do not perform account or payment actions.

AI previews are generated from your saved photo and a chosen styling instruction. They may change facial details as well as the selected look. They are illustrative possibilities, not predictions of a haircut or achieved results. Procedure simulations are unavailable.

What evidence means here

A source can explain a measurement or provide context without proving that a recommendation will help a particular person. We distinguish those claims. A photograph cannot establish a medical condition, a person’s health, or a guaranteed response to a change.

Data and consent

Optional on-device analysis runs locally without uploading a photo or requiring an account. Saving a photo for AI previews or an optional styling plan requires sign-in and separate upload consent; AI generation requires its own consent. The AI styling plan uses one front photo in private storage. Every uploaded photo and generated preview is removed within 30 days, or sooner on a deletion request. The annotations on the guide page are calculated on your device and are not stored. Read the privacy policy for retention and deletion.

Editorial responsibility and scope

FaceFluent publishes the learning guides under an organizational byline. We sell the product discussed in those guides. The byline does not represent a named stylist, clinician or independent expert review.

AI tools assist with drafting and revising these guides. FaceFluent is responsible for what it publishes; AI assistance does not establish styling expertise or independent fact-checking. Product instructions are checked against the implemented workflow, and source links identify the scope of supporting evidence.

Haircut decision questions are editorial planning suggestions. External references support only the stated technical or hair-care context; our own service pages describe product behavior. Neither type of source proves that a suggested style will suit an individual. Revised guides show an updated date alongside the original publication date.

If a guide conflicts with the current product or misstates a source, report the page and passage to support. Do not include a personal photo to report a content error.

Technical reference

Google MediaPipe Face Landmarker documentation describes the underlying landmark detector. Its capabilities are not a validation of FaceFluent’s aesthetic guidance.