Face Reading AI: Bringing Chinese Physiognomy to the Web

This browser-based face reading AI app captures a user’s facial scan or uploaded image and generates a Chinese physiognomy report using an AI-driven analysis engine. It delivers an end-to-end flow from image capture through AI interpretation to report viewing and administration.

Challenge: Face Scanning, AI Reliability, and Performance

The project faced several technical and UX challenges across image capture, AI interpretation, and performance.

• Face scan UX on Android was limited by the underlying capture library.
• Auto-capture during scanning was unreliable and inconsistent.
• Existing tools (such as MediaPipe) did not provide full landmarks for the forehead and ears, requiring prompt-based AI estimation of these regions.
• Smile-related image prediction was not sufficiently accurate.
• AI outputs were inconsistent between runs and sometimes too generic or hallucinated.
• Tone control was difficult, with results occasionally too soft or too harsh.
• Large prompts and data payloads slowed response times from the AI layer.
• Inconsistent output formats from the AI occasionally broke UI rules and layouts.

Solution: Modular Face Reading Application

To handle these issues systematically, the product was organised into four core modules that together support the complete experience.

Module 1: Face Scanning Module (Web-based)

This module focuses on getting high-quality facial input into the system.

• Capture photos using the device camera directly in the browser.
• Accept image uploads for users who prefer using existing photos.
• Preprocess facial data by extracting landmarks, ratios, and symmetry metrics.

Module 2: AI Physiognomy Engine

This is the analytical core of the app.

• Examine key facial regions such as forehead, nose, lips, jaw, and eyes.
• Compare feature patterns against Chinese physiognomy reference frameworks.
• Generate written interpretations of traits, tendencies, and insights based on the scan.

Module 3: Report Generation & Management Interface

• This module turns AI output into readable, reusable insights.
• Build visual and text-based reports from each scan or upload.
• Save reports to the user session or account for later review.
• Enable users to download or share their readings externally.

Module 4: Admin Management Portal

This module supports operational visibility and continuous improvement.

• Display a list of users and their reading history.
• Help admins monitor AI behaviour and identify cases needing refinement.

Impact: Solid Foundation and Ongoing Optimisation

The outcome is a functioning platform that covers all required modules, from scanning and AI analysis to reporting and administration, forming a robust base for a production-grade face-reading service. The team is actively integrating client feedback on AI-generated readings to refine accuracy, tone, and consistency, aligning the interpretive output more closely with client expectations over time.