Transcriber

A flexible Python package for transcribing audio and video from various sources (SoundCloud, YouTube, or local files) into multiple text formats (txt, docx, srt). The transcriber supports multiple speech recognition models (defaulting to Whisper "large-v3"). It leverages GPU acceleration for faster processing and utilizes concurrency through async and multiprocessing to improve performance.

Features

• Multiple Input Sources: Transcribe from SoundCloud, YouTube, or local audio/video files
• Multiple Models: Support for both faster-whisper and standard whisper models
• GPU Acceleration: Utilize GPU processing for faster transcription
• Concurrent Processing: Batch processing and VAD filtering for improved performance
• Various Output Formats: Export transcriptions to multiple formats
• Progress Tracking: Rich progress bars showing transcription status
• Configurable: Extensive configuration options via environment variables or settings

Requirements

• Python 3.11 or higher
• GPU support (optional but recommended for better performance)

Installation

• Install the package from PyPI:

  uv pip install Transcriber
  

Configuration

• Copy the example environment file:

  cp .env.example .env
  

• Configure the settings in .env:

  # Input settings
  INPUT__URLS_OR_PATHS=["Audio_To_Transcribe"]
  INPUT__SKIP_IF_OUTPUT_EXIST=false
  
  # Output settings
  OUTPUT__OUTPUT_FORMATS=["all"]
  OUTPUT__OUTPUT_DIR="Transcripts"
  
  # Whisper model settings
  WHISPER__LANGUAGE="ar"
  
  # Logging settings
  LOGGING__LOG_LEVEL="DEBUG"
  LOGGING__ENABLE_LOGFIRE=true
  LOGGING__LOGFIRE_TOKEN="<your_logfire_token>"
  

Usage

Basic Usage

from Transcriber.transcriber import transcribe

# Configure the transcriber via .env or parameters, then:
transcribe(
    urls_or_paths=["path/to/audio.mp3", "https://youtube.com/watch?v=example"],
    output_dir="Transcripts",
    output_formats=["txt", "srt"],
    language="ar",
    log_level="INFO",
    enable_logfire=True,  # Optional
    logfire_token="<your_logfire_token>"  # Optional, you can get it from https://logfire.pydantic.dev/login
)

Command Line Usage

uv run --with Transcriber transcribe

Example Configuration

# Example settings in your .env file
INPUT__URLS_OR_PATHS=["path/to/audio.mp3", "https://youtube.com/watch?v=example"]
OUTPUT__OUTPUT_FORMATS=["txt", "srt"]
WHISPER__LANGUAGE="en"

Acknowledgments

• OpenAI Whisper
• Faster Whisper