Darpan

Magic Mirror

Comment

Inspiration

What if a mirror could do more than just show your reflection? What if it could understand your skin, guide your skincare routine, help you experiment with different looks, and tell you whether your skincare products are actually making a difference?

Inspired by the magical mirrors found in fairy tales, we imagined Darpan — a modern-day magic mirror powered by AI.

Traditional mirrors only reflect how we look. Existing beauty applications often provide isolated features such as skin analysis or virtual makeup try-ons. We wanted to bring these experiences together into one interactive, intelligent companion.

Our vision was to create a mirror that doesn't just tell you what you want to hear, but helps you understand your skin through meaningful insights and honest feedback.

What it does

Darpan is an AI-powered talking magic mirror that transforms everyday skincare and beauty routines into an interactive experience.

Using YouCam's AI capabilities, Darpan offers:

  • AI Skin Analysis: Analyze a photograph using YouCam Skin AI and receive seven measurable skin indicators: smoothness, hydration, calm tone, radiance, oil balance, fine pores, and clarity.
  • Personalized Skincare Coach: An interactive voice-enabled assistant that guides users through morning and evening skincare routines, one step at a time.
  • Virtual Makeup Studio: Explore five different makeup looks, follow step-by-step application instructions, and preview looks on your own photograph using YouCam's virtual try-on capabilities.
  • Before & After Verification: Compare skin scans over time to understand whether there have been measurable changes, while accounting for lighting differences, measurement uncertainty, and scan timing.
  • Royal Ledger: Track skincare products across multiple scans and receive evidence-based progress assessments instead of premature claims about product effectiveness.
  • Smart Product Recognition: Use the camera to capture and read skincare product labels with AI-powered image understanding.
  • Multilingual Interaction: Interact with Darpan in English, Hindi (Devanagari), and Hinglish, through both text and voice.

Darpan is designed to feel less like a conventional application and more like having an intelligent mirror in your room.

How we built it

We built Darpan using a modern web architecture, combining computer vision, AI-powered analysis, virtual try-on technology, and conversational interaction.

Technology Stack:

  • Frontend: Next.js 16, React 19, TypeScript, and Tailwind CSS 4.
  • Skin Intelligence: YouCam API for AI-powered skin analysis and beauty-related capabilities.
  • Virtual Try-On: YouCam API for makeup look previews.
  • Conversational AI: Groq as the default LLM provider, with support for xAI Grok, Gemini, OpenRouter, and custom OpenAI-compatible endpoints.
  • Computer Vision: MediaPipe Tasks Vision for face-related visual processing.
  • Voice Interaction: Browser-based speech recognition and speech synthesis.
  • Backend: Next.js server-side API routes with validation, request protection, and controlled API usage.
  • Local Persistence: Browser local storage for skincare history, routines, and product tracking.

How the system works:

  1. The user interacts with Darpan through its camera, microphone, or text interface.
  2. With appropriate consent, an image is processed through the YouCam API to obtain skin analysis results.
  3. The backend validates and processes the results before presenting them through an intuitive visual dashboard.
  4. The conversational AI interprets user requests and provides context-aware skincare guidance.
  5. Users can track routines, compare scans, experiment with makeup looks, and monitor product progress over time.

We also implemented a dedicated Demo Mode that allows users to explore the application without API credentials. Demo measurements are explicitly labelled as samples rather than being presented as real skin analysis.

Challenges we ran into

Building Darpan involved challenges beyond simply integrating AI APIs.

  • Making AI measurements meaningful: Skin analysis produces multiple measurements, but presenting numbers without context can be misleading. We designed seven visual indicators with numerical scores and measurement bands to make the results easier to understand.
  • Avoiding misleading AI claims: A product should not automatically be considered effective simply because two scans show different numbers. We introduced uncertainty bands, scan-timing checks, and conservative verdicts to avoid overstating results.
  • Creating natural conversations: Making an AI assistant feel like a mirror rather than a chatbot required designing short, contextual responses and voice interactions that fit the skincare experience.
  • Privacy-first image processing: We wanted users to remain in control of their photographs. Images are only sent for processing after explicit user consent, and the application does not save photographs or skin-tone readings on its server.
  • Reliable API integration: Handling external API failures, rate limits, usage costs, and different response formats required dedicated validation, error handling, and spending safeguards.
  • Building a cohesive experience: Combining skin analysis, skincare routines, makeup previews, product tracking, and voice interaction into one interface without making it overwhelming was a significant design challenge.

Accomplishments that we're proud of

  • Built a functional AI-powered magic mirror experience rather than a collection of disconnected beauty tools.
  • Integrated YouCam's skin analysis and virtual try-on capabilities into a unified application.
  • Created an interactive, multilingual voice-enabled skincare companion.
  • Developed a product tracking system that prioritizes measurable progress over exaggerated AI claims.
  • Implemented privacy-first image handling and a consent-based processing workflow.
  • Built a fully explorable Demo Mode that works without external API credentials.
  • Created a distinctive visual identity inspired by traditional palace mirrors, using a jewel-inspired colour palette and a magical mirror interface.
  • Verified the application through automated unit tests, production builds, and Playwright end-to-end tests.

One of our biggest accomplishments is that Darpan is designed around a simple principle: AI should help people understand themselves better, not make unrealistic promises about how they look.

What we learned

Through this project, we learned that building an AI-powered product is not just about connecting APIs to a frontend.

We learned how important it is to:

  • Translate complex computer vision outputs into understandable experiences.
  • Design conversational AI around a specific context rather than generic chatbot interactions.
  • Account for uncertainty when presenting AI-generated measurements.
  • Treat user consent and privacy as fundamental architectural decisions.
  • Build reliable integrations with external AI services while controlling costs and handling failures.
  • Combine multiple AI capabilities into a cohesive product experience.

Most importantly, we learned that a good AI product should know when it does not have enough evidence to give an answer.

What's next for Darpan

We envision Darpan evolving from a web-based magic mirror into a more immersive personal beauty and skincare companion.

Our next steps include:

  • Physical Magic Mirror: Explore integration with smart displays and two-way mirrors to create a real-world Darpan experience.
  • Deeper Personalization: Improve recommendations using longer-term skin analysis trends, individual preferences, and skincare history.
  • Expanded Virtual Try-On: Introduce more hairstyles, makeup styles, accessories, and fashion experimentation.
  • Advanced Product Intelligence: Improve ingredient recognition and help users understand product compatibility and skincare routines.
  • More Natural Conversations: Enhance voice interaction to make conversations feel increasingly natural and context-aware.
  • Broader Accessibility: Expand multilingual support and make the experience accessible to more users.

Our ultimate vision is simple:

A mirror that doesn't just show you your reflection. It helps you understand it.

Built With

Share this project:

Updates

Submission history