Watch AI – Virtual Try-On Platform
Laravel and Google Gemini AI
Project Brief
Smart Web Agency designed and developed AI, a web-based virtual try-on platform that allows users to visualise themselves wearing different watch styles in personalised environments through AI-powered image generation. The goal was to combine e-commerce innovation with realistic AI rendering to deliver an immersive try-before-you-buy experience for watch enthusiasts across the UK. Security and compliance were priorities throughout. We applied high practice standards including encryption, role based access, data protection and secure hosting on our managed servers. With continuous monitoring and regular updates, the system runs reliably with minimal downtime.
Core Functionalities
Virtual Try-On
Users upload a personal photo, choose from three watch categories like Sport, Luxury, and Casual, and describe a custom environment. Watch AI then generates a photorealistic image showing the user wearing the chosen watch in that setting.
Intelligent Image Generation
- Identity Preservation: Facial features, body type, glasses, and skin tone remain intact.
- Dynamic Scene Customisation: Backgrounds, outfits, and lighting adapt to the prompt while keeping realism.
- High-Resolution Output: Each image is generated at 2048×2048 pixels or higher.
Usage & Rate Limiting
- 20 generations per hour per IP address.
- Real-time counter display with visual status colours (green → yellow → red).
- Ensures fair usage and cost control.
Image Management
- Chat-style interface for interactive results.
- Full-screen lightbox gallery with downloads.
- Session persistence: user uploads remain available for subsequent generations.
Admin Panel
A secure admin dashboard allows SWA and the client to:
- View all generated images in a grid gallery.
- Browse with keyboard navigation.
- Inspect metadata such as watch type, timestamp, and session ID.
Technology used
Backend
Framework: Laravel 12 (PHP 8.2)
Database: MySQL
Caching: Laravel Cache for rate limiting
Sessions: Database-backed
Frontend
Styling: Tailwind CSS 4.0
AI Integration
Model: Google Gemini 2.5 Flash Image Preview
API: REST-based generative AI endpoint
Encoding: Base64 for efficient image handling
How It Works
Upload
Users add a photo (drag-and-drop or file upload).
Configuration
They select a watch type and write a brief scene description (minimum 10 characters).
AI Processing
The backend validates inputs, checks rate limits, and creates a database record. The Gemini API receives both user and watch images (encoded in Base64) along with a detailed prompt.
Status Polling
The frontend checks generation progress every 2 seconds, showing a live loader until the result appears.
Result Handling
The finished image is stored, marked complete, and displayed instantly in the chat view, ready to download or share via unique UUID links.
Results
Key Technical Decisions
Session-based architecture for instant use without registration.
IP-level rate limiting to prevent abuse and manage API costs.
Synchronous AI calls for real-time feedback.
Mandatory prompts to ensure contextually rich results.
Progressive enhancement — basic uploads work even if JavaScript is disabled.
Performance and Scalability
Average generation time
5–15 seconds
Image size
1–3 MB PNG
Rate limit
20 generations/hour (sliding window)
Supports
Multiple concurrent sessions without delay
Value Delivered
For Users
Realistic visualisation before purchase
Instant, account-free experience
Easy sharing of generated styles
Consistent identity preservation
For Businesses
Enhanced customer engagement and conversion rates
Insightful analytics from admin panel
Cost-efficient AI generation using Gemini Flash
Scalable architecture ready for multi-brand deployment
Results
5× faster engagement compared with static watch catalogues
High user retention due to interactive design
Seamless AI workflow delivering 2048×2048 outputs within 15 seconds
Zero downtime under concurrent use
Conclusion
The Watch AI Virtual Try-On Platform demonstrates Smart Web Agency’s ability to merge AI innovation with practical UX for retail. By combining Laravel 12, Tailwind CSS, and Gemini Flash AI, we built a secure, high-performance system that delivers premium visual experiences to UK watch retailers and customers alike. This project reflects our continued commitment to creating intelligent, scalable digital products that redefine e-commerce engagement. The result is a dependable, modern platform that streamlines dispatch, reduces manual work and supports business growth for one of Greater London’s leading taxi firms.
Frequently Asked Questions (FAQs)
A virtual try-on platform uses AI and image generation technology to let users upload a photo and visualise products—like watches, clothing or accessories—on themselves. The system analyses the photo, applies the selected product, and produces a realistic image showing how it would look in real life
AI enhances realism by preserving facial and body details while intelligently adapting the background, lighting and outfit to match the chosen product and scene. This creates lifelike visuals that help users make confident purchase decisions.
AI try-on technology increases engagement, reduces product returns, builds customer trust and personalises the shopping experience. It allows UK retailers to stand out with immersive, interactive online experiences.
Watch AI uses advanced face- and body-preservation techniques so the user’s facial features, skin tone, and body proportions remain identical. Only the environment and clothing adapt to the scene prompt.
Yes. Watch AI outputs high-resolution images at a minimum of 2048×2048 pixels, suitable for digital campaigns, product previews, or personal sharing.
Each user (or IP address) can generate up to 20 images per hour. A dynamic counter displays remaining generations using green, yellow, and red indicators to maintain fairness and server efficiency.
No sign-up is required. The system runs on session-based logic, allowing users to upload, generate and download images instantly without registration.
Businesses gain valuable engagement analytics, conversion data, and usage insights through the admin dashboard. The tool reduces hesitation to buy and improves marketing visuals at minimal cost.
Watch AI is built on Laravel 12 and integrates with Google Gemini 2.5 Flash for image generation. It uses Tailwind CSS, jQuery, and SQLite, with secure caching, storage, and logging handled by Laravel’s native services.
Upcoming features include live Augmented Reality (AR) try-on, multi-product scene generation, speech-to-prompt interaction, and deeper analytics for e-commerce conversion tracking.