VerbaSense
March – April 2026
Team project · Internship team project at USSBD, supervised by Md. Harun Or Rashid, Senior Software Developer, who guided architecture and reviewed the code. It is an internal USSBD project; the repository's commits are made from my account, but the product was built under the mentor's direction and not independently.
Overview
VerbaSense is a speech-to-text web application built during my software engineering internship at Unicorn Software & Solutions Ltd (USSBD). Users record or upload audio, choose a language, and receive a transcription produced by the whisper.cpp engine, with a saved history they can review, edit, copy and export.
- Problem
- Speech-to-text tools are often tied to third-party transcription APIs and built around English. The goal was a transcription product that runs the Whisper engine on its own server, handles several languages, and keeps each user's transcription history in their own account.
- Solution
- The API converts each recording with FFmpeg and transcribes it with the whisper.cpp command-line engine, using an English model for English and a multilingual model for other languages. A Next.js dashboard handles recording, upload, language selection and history, and an Express API with JWT authentication and MongoDB stores every user's transcriptions.
Key features
- Record audio in the browser or upload a file (WAV, MP3, M4A, OGG, FLAC or WebM)
- Language selection (English, Hindi, Spanish, Arabic, Marathi or auto-detect in the interface); English uses a dedicated model and other languages a multilingual one
- Transcription history with duration and language: view, edit, copy, download as a .txt file, delete one entry or clear everything, with confirmation dialogs
- Email and password sign-up with email verification, Google sign-in, and password reset by email
- Profile page with photo upload, and a contact form that confirms the sender's email with a verification code
- Light and dark themes with a responsive interface
Tech stack
- Next.js
- React
- Tailwind CSS
- shadcn/ui
- Framer Motion
- TypeScript
- Node.js
- Express.js
- MongoDB
- Mongoose
- JWT
- whisper.cpp
- FFmpeg
- Cloudinary
Architecture & technical details
- Next.js client
- Express API
- FFmpeg + whisper.cpp
My role & contribution
Role: Software engineering intern (full-stack)
- Engineered the REST API modules for authentication, transcription and history, including JWT authentication, email verification, Google sign-in and password reset
- Integrated whisper.cpp and FFmpeg into the API for audio processing
- Built the responsive Next.js interface: the dashboard, recording and upload, language selection and transcription history
- Took part in requirement analysis, debugging, performance and architecture discussions with the mentor