AI Selfie App 2.0: Find yourself in a story

This project presents the development of the AI Selfie App 2.0, a web-based application that allows visitors to generate AI portraits of themselves.

Initial situation

AI Selfie App is an interactive web application that lets visitors take a selfie, transform it into an AI-generated portrait, and see themselves as part of a unique story. Designed for public exhibitions, it combines a fast and intuitive user flow with clear consent handling, fun personalization options, and an educational, step-by-step explanation of how AI image generation works. The app runs entirely in the browser, requires no account, and is optimized for short, engaging sessions suitable for both children and adults.

Problem statement / Project goal

The goal was to adapt and enhance the AI Selfie App for real-world exhibitions by:

• Ensuring smooth, short-session usability.
• Implementing an explicit but lightweight consent flow.
• Providing clear, narrative explanations of AI image generation.
• Optimizing the UI and UX for public use.

Technologies Used

The AI Selfie App is built using React Native and the Expo framework, enabling its deployment as a cross-platform web application. Supabase provides a robust backend solution for database management, user authentication, and secure file storage. The generative AI model itself is accessed via a RESTful API, and the application's modern frontend design is crafted with Tailwind CSS.

• Frontend: React Native Web + Expo, Tailwind/Nativewind
• Backend: Supabase (PostgreSQL, Storage, Auth, Cron), serverless API routes for AI inference requests
• AI Processing: Stable Diffusion v1.5 fine-tuned on curated face/style datasets
• Deployment: Expo EAS Hosting for single-command rollout to exhibition devices

Solution developed and its benefits

The AI Selfie App offers a unique and interactive experience with several notable features:

• Web-Based Access: Runs in the browser, no installation required.
• Anonymous Sign-In: Generates unique IDs without collecting personally identifiable information.
• Personal AI Portrait Generation: Upload a selfie to generate AI-created portraits in various artistic styles.
• Privacy-Preserving Consent: Interactive consent mechanism ensures no image is uploaded without explicit agreement.
• Explainable AI "Wizard": Step-by-step visual and narrative explanation of Stable Diffusion, inspired by story-driven games.
• Sample Profile: Allows exploration without uploading personal images.
• Daily Data Cleanup: Automatically deletes all uploaded and generated images older than 24 hours.

The development followed an iterative, research-informed approach, including literature review, usability testing with think-aloud protocols, and standardized measurement tools such as the User Engagement Scale (Short Form) (UES-SF) and the System Causability Scale (SCS). Agile principles guided the process.

Results and Impact

The AI Selfie App provides a functional and engaging framework for playful interaction with AI-generated imagery while addressing key ethical and educational considerations. Although further improvements are planned, the current state of the application is ready for deployment and use in its intended exhibition context

Key terms

• AI Selfie App
• Web Application
• React Native
• Supabase
• Artificial Intelligence (AI)
• Generative AI
• Explainable AI (XAI)
• Usability Testing

FHNW – Institute of Interactive Technologies (IIT)
Hochschule für Technik FHNW
5210 Windisch
Institute website

Team