A browser-based application that captures microphone input, transcribes speech using Whisper AI, and translates English to Hindi in real-time. Fully containerized with Docker for easy deployment.
- 🎤 Live microphone recording via browser
- ✍️ Accurate transcription using Whisper (tiny model)
- 🌍 English → Hindi translation via Argos Translate
- ⚡ 5-second processing cycles for near real-time results
- 🐳 Full Docker integration (frontend + backend)
- 🚀 Modern stack (React, FastAPI, Tailwind CSS)
- Docker & Docker Compose
- Git (optional: Git LFS for large model files)
git clone https://github.com/AkankshRakesh/realtime-transcriber.git
cd realtime-transcriberDownload the English-Hindi model from Argos Open Tech and place it in the backend folder:
mv translate-en_hi-1_1.argosmodel backend/translate-en_hi.argosmodelNote: If the model exceeds 100MB, consider using Git LFS.
docker-compose up --buildAccess the application: Frontend: http://localhost:5173 Backend API: http://localhost:8000
POST /transcribe| Content-Type | File | Response |
|---|---|---|
multipart/form-data |
recording.webm |
transcript, translation |
🏗️ Project Structure
.
├── backend/
│ ├── main.py # FastAPI application
│ ├── Dockerfile # Backend container config
│ ├── requirements.txt # Python dependencies
│ └── translate-en_hi.argosmodel # Translation model
├── frontend/
│ ├── src/components/ # React UI components
│ ├── Dockerfile # Frontend container config
│ ├── package.json
│ └── tailwind.config.js # Tailwind CSS config
└── docker-compose.yml # Multi-container orchestration
Frontend: React, Tailwind CSS, Vite
Backend: FastAPI, Python
Transcription: faster-whisper
Translation: Argos Translate
Audio Processing: FFmpeg
Containerization: Docker, Docker Compose