70%
People often miss the part of communication that matters most: tone, hesitation, confidence, urgency, and subtext.
A sentence can sound polite and still be tense. It can sound clear and still hide doubt.
About the project
Built to understand people, not just words.
ToneLens is a real-time emotional intelligence system for multilingual conversations. It combines Google Gemini Live, FastAPI, Cloud Run, Firestore, and Vertex AI to help people read tone, intent, and context as the conversation happens.
The Problem
Most systems translate text. Very few help people understand what someone actually means in the moment.
Why ToneLens exists
The project was built around a simple gap. Translation tools help with words, but they do not help with emotional context or intent.
ToneLens turns live speech into a clearer decision surface by pairing transcription, multimodal reasoning, and lightweight session memory. The goal is not to replace human judgment, but to make it easier to trust what you are hearing.
The Builder
A clean snapshot of the person behind the build, focused on clarity over theatrics.
Mohan Prasath
18 years old, Chennai, B.Tech CSE AI&ML
I build systems that feel practical, fast, and usable. ToneLens is the result of combining AI, product thinking, and a strong bias toward simple interfaces.
- Age: 18
- Location: Chennai
- Education: B.Tech CSE AI&ML
Tech Stack
Five pieces that keep the app responsive, deployable, and understandable end to end.
01
Google Gemini Live API
Real-time multimodal reasoning and conversation understanding.
02
Cloud Run
Serverless deployment with production-ready scaling and reliability.
03
Firestore
Session memory and lightweight state persistence across turns.
04
FastAPI
A clean Python backend for orchestration, routing, and app services.
05
Vertex AI
Response refinement and structured output shaping where needed.
Build Timeline
Seven days of iteration, tightening the product from core concept to polished demo.
Day 1
Project skeleton and first integration
The app structure, backend entry points, and the first working end-to-end flow were set up.
Day 2
Conversation and UI foundation
The early interface, conversation handling, and core screen experience were shaped.
Day 3
Structured output and formatting
The response pipeline was tightened so the assistant could produce clearer, more useful output.
Day 4
Bridge refactor and routing cleanup
Backend coordination improved, with cleaner orchestration between model calls and app state.
Day 5
Frontend polish and layout tightening
The interface was simplified, spacing was refined, and the product started to feel coherent.
Day 6
Responsive fixes and reliability work
Mobile behavior, buffering, logging, and stability improvements were added across the stack.
Day 7
Final submission polish
Design cleanup, footer updates, and the final launch pass brought the project across the finish line.
What's Next
A short roadmap focused on making ToneLens more useful in more places.
- Mobile app for a lighter on-the-go experience
- Browser extension for faster conversation support
- Enterprise API for integrations and private deployments
- Real-time wearable or earpiece mode for live assistance
- ✓ Gemini Live Agent Challenge 2026 submission