1. What we measure
Every SoftMaxx scan returns a numeric breakdown across 11 evaluation pillars. Each pillar is scored from 0 to 10, and the overall score is a weighted aggregate.
Pro and Elite scans add a side-profile cephalometric analysis that evaluates additional measurements not visible from the front:
Each pillar comes with a short prose comment explaining the rating, plus a ranked list of which pillars have the most room for improvement.
2. Research basis
Our scoring rubrics are calibrated against frameworks drawn from orthodontic and maxillofacial research literature. Specifically:
- Phi (golden) ratio proportions — the 1.618:1 relationship that appears repeatedly in facial-aesthetic literature.
- Ricketts E-line — a line drawn from the nose tip to the chin point, used to evaluate lip projection in profile view.
- Soft-tissue analysis — standard cephalometric reference points (nasion, subnasale, pogonion, etc.) used in orthodontic planning.
- Facial thirds and fifths — horizontal and vertical proportionality benchmarks established in classical anatomy.
These are research frameworks, not absolute laws. Different cultures, sexes, and eras weight facial features differently, and academic literature itself does not converge on a single canon of beauty.
Important caveat: AI vision models calibrated against research literature do not replace clinical evaluation. SoftMaxx is an analytical tool, not a diagnostic one.
3. How the AI does it
The scoring pipeline runs in three stages:
- Landmark extraction. Your photo is processed in your browser by MediaPipe (an open-source library from Google) to extract 478 anatomical landmark points. These points are used to compute geometric ratios (jaw width, eye spacing, midface-to-lower-face proportion, etc.).
- Vision analysis. Your photo and the geometric ratios are submitted to OpenAI's vision API, which evaluates your face against the curated rubric described above and returns a structured score.
- Protocol generation. Based on your score breakdown, the AI generates a ranked list of recommended improvements and a personalized routine. Pro and Elite tiers receive more nuanced protocols.
Model selection by tier
| Tier | Model | Tradeoff |
|---|---|---|
| Free | gpt-4o-mini | Faster response, less detailed nuance. |
| Pro | gpt-4o | More nuanced ratings and richer protocol commentary. |
| Elite | gpt-4o | Same model as Pro, plus PSL market-position and golden-ratio analysis. |
Confidence reporting
Every scan reports an AI confidence level — HIGH, MEDIUM, or LOW — that reflects how clearly the AI could read your photo. Confidence depends on lighting, angle, expression, and image resolution. A low-confidence scan is not wrong; it is uncertain. If you receive a LOW confidence rating, we recommend a rescan under better conditions.
4. What it cannot do
Not a medical diagnosis. SoftMaxx does not diagnose skin conditions, dental malocclusion, asymmetry pathology, or any other medical condition. For clinical evaluation, consult a board-certified plastic surgeon, dermatologist, or orthodontist.
- Not surgical advice. Score breakdowns are aesthetic observations, not procedure recommendations. Do not use SoftMaxx output to plan surgery.
- Beauty is subjective. Our scoring captures one analytical lens calibrated against published proportionality research. It is not absolute truth, and reasonable people weight features differently.
- Photo quality affects accuracy. Bad lighting, off-axis angles, exaggerated expressions, heavy filters, or low-resolution images reduce the AI's confidence and can shift scores. Use a neutral expression, even front lighting, and a phone-quality camera for best results.
- The AI can be wrong. Vision models occasionally misread features. If a rating feels obviously off, the confidence flag is a useful signal; consider a rescan.
5. How we handle your data
Your photo is processed transiently. Specifically:
- Photos are not retained on our servers after analysis completes — typically destroyed within seconds.
- The 478 landmark points are computed in your browser and sent to our backend as numeric ratios, not as a face template usable for identification.
- We are BIPA-compliant (and observe the Texas CUBI and Washington 19.375 biometric privacy frameworks) for Illinois, Texas, and Washington residents.
- Your scan history is stored in your browser's localStorage, not on our infrastructure.
For full details, see our privacy policy.
6. Why a subscription, not a one-time charge?
An honest answer: a one-time scan is a single data point. A subscription is a tracking system.
SoftMaxx is designed around the assumption that you'll rescan over time — after a month of consistent skincare, after a haircut, after dental work, after a glow-up protocol. Each rescan generates a delta against your prior scan, and the AI chat lets you ask follow-up questions about your protocol as you progress.
You get a one-time scan for free, with no signup required. The subscription unlocks the side-profile / cephalometric layer, the full protocol, the rescan history, and ongoing AI chat as your face changes.
Pro is $19.99/month. Elite is $39.99/month or $150/year. Both cancel anytime with a 30-day money-back guarantee on first purchase.
If a subscription model isn't right for you, the free tier remains useful as a snapshot — nothing about SoftMaxx is paywalled to the point of unusable. See our refund policy for details.