Inspiration
Most emotion detection tools only see what people show. I wanted to build something that could detect what people hide. During 11th grade, I noticed how much emotional suppression was happening around me — students smiling through anxiety, staying silent when frustrated, and pretending everything was fine. I was doing the same. While researching for a school project, I discovered Rosalind Picard’s work on affective computing. That became the spark. I wanted to create a private, judgment-free tool that could gently reveal hidden and suppressed emotions to help people understand themselves better.
What it does
SubtleSense is an AI-powered emotion intelligence app that helps users understand not just what they show, but what they hide and suppress. Free users can do basic emotion analysis. After signing in, users unlock exclusive features: personalized moodboard, emotion timeline & graphs, a private Journal for self-reflection, weekly emotional reports, dark/light mode, and profile customization.
How I built it
I built SubtleSense as a solo project. I started with Lovable (an AI app builder) to quickly turn my idea into a working web prototype. Later, I converted the web app into an Android app using Median.co so more people could use it easily. The core uses Gemini AI for multimodal analysis (facial landmarks, micro-expressions, and voice tone). I designed the UI myself, focused heavily on privacy, and kept iterating based on real user feedback.
Challenges
The biggest technical challenge was the “Failed to Detect” error caused by JSON parsing issues (receiving HTML instead of valid JSON from the API). Debugging this while managing 12th grade studies, SAT prep, and family expectations was exhausting. On the personal side, I struggled with low signup rates initially, emotional overwhelm, and self-doubt. There were days I cried and wanted to quit. Balancing this project with boards and SAT while dealing with family pressure tested my emotional strength a lot.
Accomplishments that we're proud of
-Built a fully functional Emotion AI app from scratch as a 16-year-old high school student. -Achieved 670+ visitors with an average session time of 6 minutes. -Successfully converted the web app into an Android app. -Maintained strong privacy (client-side processing, no video storage). -Created a tool that actually helps people see their hidden emotions through moodboard and personalized advice.
What’s next
Official Android launch on May 16, 2026. After launch, focus on improving retention through the Journal and weekly reports, and exploring ways to make emotional self-awareness even more accessible.
Built With
- css
- gemini
- pwa
- react
- superbase
- taliwind
- typescript
- vite
Log in or sign up for Devpost to join the conversation.