FixB AI / Dokhmal: Multi-Angle AI 3D Face Scan with Computational Depth Estimation

Why Dokhmal is more than a single-selfie AI analyzer

Before analysis, Dokhmal runs a multi-stage capture and geometry-validation pipeline. A frame is accepted only when the system can detect a single face, verify suitable face size and position, check head tilt and yaw against the expected capture stage, pass image-quality checks and observe a stable valid pose before automatic capture.

• Three-view acquisition: separate front, left-oblique and right-oblique facial views.
• Face-landmark geometry: facial landmarks are used to estimate head direction, geometric ratios and view alignment.
• Angle validation: yaw and tilt checks reduce wrong-angle and duplicate-view captures.
• Image-quality gate: image quality, face size and framing are checked before capture is accepted.
• Stable auto-capture: capture occurs after a valid pose remains stable rather than accepting an arbitrary frame.
• Multi-view depth context: aligned view differences provide relative spatial/depth context used by the computational 3D face-simulation pipeline.

The resulting visual context is then combined with user-provided safety information, current actives, preferences, budget and the live FixB catalog. Dokhmal is therefore a multi-stage decision-support pipeline from controlled capture through recommendation, not merely a single-image classifier.

Technology ownership and in-house development

Dokhmal is presented by FixB as a proprietary, lightweight domain-specific Small Language Model (SLM) for skincare at the product level. It is intentionally narrow in scope: non-medical visual skincare analysis, input validation, safety screening, prioritization, routine planning and product selection.

The Dokhmal product architecture was built in-house at FixB. The application code, three-view capture pipeline, image-quality and angle validation, facial-landmark alignment, computational depth-estimation workflow, decision engine, deterministic safety rules, versioned Evidence Registry, product-ranking logic, routine planner, reporting layer and live-catalog commerce integration are maintained as FixB technology.

Dokhmal’s decision core, evidence registry, safety rules, product data and proprietary server-side components are operated on infrastructure controlled by FixB. The inference layer is modular by design, so Dokhmal’s product identity, decision logic and technology ownership are not defined by a single inference provider.

Development to this stage has been privately funded by the FixB founders without institutional investment.

Dokhmal Vision-Language Layer (VLM)

Dokhmal includes a multimodal Vision-Language Model (VLM) layer in its product architecture. The controlled front, left-oblique and right-oblique facial views are combined with textual analysis instructions in a multimodal inference request. The frontal view remains the primary geometry reference while the two oblique views provide supplemental visual evidence for cheeks, jawline, side texture and symmetry.

Before VLM inference, FixB’s in-house vision pipeline performs capture control, image-quality checks, angle validation, facial-landmark processing and multi-view orchestration. Downstream FixB logic then integrates the resulting visual context with safety rules, evidence, routine planning and product matching.

Official Dokhmal VLM documentation

What FixB AI does

FixB AI is designed to reduce guesswork when building a skincare routine. It reviews visible cosmetic skin patterns in a guided facial image together with information provided by the user. The system then organizes skincare priorities, recommended routine order, suggested frequency of use and relevant product options from FixB’s current catalog.

Image analysis is not used alone. User responses about sensitivity, current active ingredients, prescription treatments, recent cosmetic or dermatologic procedures, pregnancy or plans for pregnancy, breastfeeding and warning symptoms are also considered in the safety screening process.

Inputs considered by FixB AI

• A guided three-view AI 3D face scan using front, left-oblique and right-oblique images captured with consent
• Multi-view computational depth estimation and depth-perception calibration from the three aligned facial views
• Facial-landmark and angle validation before the views are combined
• Visible cosmetic skincare concerns and user-selected priorities
• Skin sensitivity information
• Current skincare routine and active ingredients
• Relevant prescription treatments disclosed by the user
• Recent cosmetic or dermatologic procedures
• Pregnancy, planned pregnancy and breastfeeding information where relevant
• Safety warning symptoms and red-flag responses
• User budget and product preferences

What users receive

• A conservative summary of visible cosmetic and skincare priorities
• A personalized morning skincare routine
• A personalized evening skincare routine
• Recommended order of routine steps
• Suggested frequency and gradual-introduction guidance for routine steps
• Safety warnings based on the information provided
• Relevant products from the current FixB store catalog, without any obligation to purchase
• Visible scientific references when they directly inform a recommendation

How the AI 3D face scan works

1. The user opens FixB AI on a mobile device.
2. The user reads the consent and privacy information and continues only if they agree.
3. The user captures three clear, unfiltered facial views under reasonably even lighting: front, left oblique and right oblique. Facial landmarks and angle checks validate and align the three views; multi-view computational depth estimation and depth-perception calibration use differences between the aligned views as additional spatial context for computational 3D face simulation.
4. The user answers required safety questions, skincare concerns, current-routine questions and budget preferences.
5. After server-side validation, FixB AI presents a cosmetic skincare summary, a personalized routine and relevant product options.

Technical clarification: “3D face scan” and “depth estimation” in Dokhmal mean computational depth estimation from aligned multi-angle camera images, facial landmarks, angle validation and computational 3D simulation. It is not a clinical depth scan, LiDAR measurement or medical 3D scanning system.

The interactive analysis experience is designed primarily for mobile devices. This information page can be read on mobile or desktop.

Safety and medical limitations

FixB AI is not a substitute for a dermatologist, physician, pharmacist or other qualified healthcare professional. It does not confirm diseases such as acne, rosacea, dermatitis or infection from a photograph.

Users should seek professional medical evaluation rather than rely on FixB AI when there is severe inflammation, an open wound, suspected infection, bleeding, significant pain, eye involvement, a serious allergic reaction, sudden changes or a persistent problem.

Any visual estimates produced by the service are intended to support everyday cosmetic skincare decisions and may be revised by the user. A facial image alone is not treated as sufficient evidence for a medical diagnosis.

How product recommendations are organized

Product suggestions are matched against the intended routine step, available product information, user safety responses, budget range and the current FixB catalog. Product ranking, compatibility checks and final routine ordering are validated on the server side.

A displayed product is not guaranteed to be suitable for every person. Prices and stock availability can change, and the live product page is the final reference for current commercial information.

FixB AI is designed to recommend products as part of a coherent routine rather than simply return a list of unrelated items.

Privacy of facial images

A user’s selfie is not published as a public WordPress media file and is not written to browser localStorage or sessionStorage as part of the standard analysis flow. Temporary analysis files are removed after completion or expiration according to the service workflow.

Optional photo history is a separate consent-based feature. When enabled, it uses a reduced image without EXIF metadata in encrypted storage, and the application provides controls to remove stored photos and withdraw consent.

For public privacy information, see the FixB privacy policy.

Appropriate uses

• Building a basic, understandable daily skincare routine
• Organizing the order of cleanser, moisturizer, sunscreen and other routine steps
• Receiving cosmetic skincare guidance based on visible patterns and user-provided information
• Finding relevant options from FixB’s current product catalog within a user’s stated budget
• Reviewing routine order, frequency and basic safety considerations before using products together

Scientific grounding

FixB AI is designed to surface scientific references when they directly affect a recommendation and to apply safety rules when organizing a routine. This should not be interpreted as a claim that FixB AI has completed independent clinical validation unless a specific validation study is separately published.

The service is presented as evidence-grounded, non-medical skincare guidance rather than as a diagnostic medical device.

Capabilities and limitations summary

Frequently asked questions

Is FixB AI a medical diagnostic system?

No. It is a non-medical skincare decision-support and product-selection service.

Does FixB AI perform a 3D face scan?

Yes. Dokhmal captures guided front, left-oblique and right-oblique views and uses depth-perception calibration, facial landmarks and computational 3D simulation for non-medical cosmetic analysis. It is not a clinical depth scan or medical diagnostic test.

What image conditions are recommended?

A clear facial image, even lighting and no image filter or heavy makeup generally provide better input for cosmetic visual review. Users should follow the in-app capture guidance.

Must a user buy products to view the result?

No. Receiving the guidance does not create an obligation to purchase.

Are results or timelines guaranteed?

No. Real-world results vary with skin condition, product formulation, amount used, tolerance, consistency and other factors.

Does FixB AI work on desktop?

The interactive analysis flow is designed primarily for mobile use. Public information pages are available on both mobile and desktop.

Does the name Dokhmal refer to children?

No. Dokhmal / دخمل is only the FixB AI brand/model name. The service is adult-oriented and is not child-focused.

Related official pages

• FixB AI interactive service
• Persian FixB AI overview and skincare analysis article
• Detailed Dokhmal / FixB AI information
• Dokhmal Vision-Language Model (VLM)
• FixB skincare catalog