## Inspiration We’ve all had that “Wait, what did they say again?” moment right after a conversation. Whether it’s a lecture, meeting, or deep 2AM heart-to-heart, the human brain just... forgets. So I thought — what if we had an app that remembered for us?
EchoTwin was born out of a need to capture, transcribe, and summarize real-life conversations — kind of like having an AI-powered brain twin. The goal? Let people live in the moment without losing the important stuff.
##What it does EchoTwin is a mobile-first app that:
Records real-world conversations (through your phone mic)
Transcribes them using Whisper AI
Summarizes the key points with Bolt AI / GPT
Delivers clean, readable summaries back to the user
Basically, it’s like Notes app x AI x Superpowers. Perfect for students, interviewers, journalists, or anyone with a forgetful goldfish memory
How we built it
Frontend: Built in React Native with Expo for easy cross-platform development
Backend: Node.js + Express hosted on Render
Audio Uploads: Handled using Multer
Transcription: Powered by OpenAI Whisper
Summarization: Generated with Bolt AI / GPT
APIs: REST architecture to connect frontend ↔ backend ↔ AI
Flow goes like this:
User records audio → file gets sent to backend
Backend runs Whisper for transcription
Transcript is passed to GPT/Bolt for summarization
Summary is returned to the user in the app
Challenges we ran into
Expo Audio Recording: Permissions, file formats, and file paths were chaotic at first
Large file uploads: Whisper sometimes choked on longer recordings → had to limit durations and compress files
CORS issues: Classic dev rite of passage. Fixed with good ol’ cors() and some backend tweaks
API rate limits + latency: Managing AI response time while keeping the app responsive
Debugging on physical devices: Logging anything was a headache. Console logs and caffeine saved the day.
Accomplishments that we're proud of Got a full AI pipeline working: record → transcribe → summarize
Built a smooth mobile interface that actually feels like a real app
Learned to wrangle audio files, deploy backend, and integrate multiple AI tools — solo.
Created a project that solves a real, relatable problem.
##What we learned How to build a full-stack AI-powered app from scratch
Real-world Expo and React Native development tricks
Managing backend file uploads and async AI calls
That Whisper is powerful, but can be picky
That deploying to Render is way easier than it looks (and free-ish!)
What's next for EchoTwin
Add user authentication and secure transcript storage
Improve the UX/UI for vibe and accessibility
Let users edit, delete, or bookmark key transcripts
Build a custom AI model for more tailored summaries
Add language support for non-English convos
Possibly integrate a “Voice Highlights” feature — like Spotify Wrapped, but for your convos 👀
Log in or sign up for Devpost to join the conversation.