Interpreto

Interpreto is a system for transcribing audio and video files. It provides an easy-to-use web interface for uploading media files and automatically generates transcriptions with timestamps. The modular architecture of the project, makes easy to add more functionality, like another interface (for example, a Telegram bot).

Interface

Features

• Upload audio and video files for transcription
• Real-time transcription updates via server-sent events
• Display of transcription alongside media player
• Automatic subtitle generation in WebVTT format
• Language detection for transcripts
• Efficient audio processing with voice activity detection

Architecture

Interpreto is built using a microservices architecture:

1. Frontend: React-based web interface for uploading files and viewing transcriptions
2. Backend API: FastAPI service that handles file uploads and job management
3. Worker Service: Processes media files using Whisper models and voice detection
4. Storage:
   • MongoDB for storing file metadata and transcriptions
   • MinIO for storing media files
   • Redis for real-time communication between services

Installation

Prerequisites

• Docker and Docker Compose
• Python 3.8+
• Node.js 14+
• CUDA-capable GPU (optional, for faster transcription)

Setup

1. Clone the repository:

git clone https://github.com/mperalsapa/interpreto.git
cd interpreto

2. Set up the environment variables (or use the defaults in the code):

MONGO_HOST=localhost
MONGO_PORT=27017
MONGO_USER=frontend-service
MONGO_PASS=frontend-service
MONGO_DB=media_service

# MinIO config
MINIO_HOST=localhost
MINIO_PORT=9000
MINIO_ACCESS_KEY=minioadmin
MINIO_SECRET_KEY=minioadmin

# Redis config
REDIS_HOST=localhost
REDIS_PORT=6379

3. Start the services using Docker Compose:

docker-compose up -d

Usage

1. Open the web application in your browser at http://localhost:8080
2. Upload an audio or video file using the upload form
3. Wait for the transcription to process - you'll see real-time updates
4. View the transcription alongside the media player
5. The transcription will be automatically displayed as subtitles

Technologies

• Frontend: React
• Backend: FastAPI (Python)
• Speech Recognition: Whisper, faster-whisper
• Voice Activity Detection: Silero VAD
• Storage: MongoDB, MinIO, Redis
• Processing: PyTorch, torchaudio