Eunoia
An AI emotional support app that hears how you sound, not just what you say, and answers with an empathetic voice.
- Client
- Final year project, COMSATS
- Year
- 2024 to 2025
- Role
- Product design, Mobile development, ML integration
- Stack
- React Native, Speech emotion recognition, NLP, TTS
Eunoia was my final year project: an app that delivers immediate emotional support through speech and text for people who cannot reach professional help in the moment.
Most chatbots read words. Eunoia also listens to tone. Speech emotion recognition classifies how the user sounds, and that signal shapes both the content and the synthesized voice of the response.
A conversation turn
- 01User speaks or types
- 02Emotion recognized from voice and text
- 03Response generated for that emotional state
- 04Spoken back with matching tone
Designing for someone having a bad day
The interface had one rule: never add friction to a person who is already struggling. One tap to talk, large calm surfaces, no streaks or gamification, and conversation history that reads like a journal rather than a chat log.
The hardest design work was tone calibration: an empathetic response that arrives too cheerful is worse than none. We tuned response style against emotion classes rather than shipping one voice for every state.
An emotional check-in, not a test
Eunoia builds in the K10, a short, validated well-being scale. Ten questions sit on a calm Always-to-Never range, the wording stays plain, and the result lands as encouragement and context rather than a clinical score thrown back at someone who is already struggling.
