Speech Transcription API

A FastAPI-based REST API service for speech-to-text transcription using NVIDIA's parakeet-tdt-0.6b-v2 model. This API provides high-quality English speech recognition with automatic punctuation, capitalization, and accurate word-level timestamps.

GitHub Repo: https://github.com/viktor2077/parakeet-tdt-0.6b-v2

Features

• 🎤 High-Quality Transcription: Uses NVIDIA's 600M parameter parakeet-tdt-0.6b-v2 model
• ⏱️ Accurate Timestamps: Provides word-level timing information
• 📝 Multiple Output Formats: JSON response or SRT subtitle format
• 🔧 Automatic Audio Processing: Handles resampling and channel conversion
• 🚀 Long Audio Support: Optimized settings for audio longer than 8 minutes
• 📊 OpenAPI Compatible: Full Swagger/OpenAPI documentation
• 🛡️ Error Handling: Comprehensive error handling and validation