Audio Transcriber

Extract text and transcribe audio from PowerPoint presentations, MP4 videos, and MP3 files using Whisper.

Features

• Text Extraction: Extracts all text content from PowerPoint slides
• Audio Transcription: Uses faster-whisper to transcribe audio from multiple sources:
  • PowerPoint files (.pptx) - embedded audio recordings
  • Video files (.mp4) - audio track extraction
  • Audio files (.mp3) - direct transcription
• Checkpoint/Resume Support: Automatically saves progress during long transcriptions
  • Resume from where you left off if interrupted
  • Checkpoints saved every 10 segments
  • Safe to stop with Ctrl+C anytime
• Live Progress Tracking: Real-time progress bar and live checkpoint file
  • View transcription progress in output/[filename]_checkpoint.json
  • See completed segments as they're processed
  • Track timestamp and text in real-time
• GPU and CPU Support: Automatic device detection with intelligent fallback
• Multiple Models: Supports various Whisper model sizes (tiny, base, small, medium, large)
• Configurable: Easy-to-modify settings for performance and quality tuning

Requirements

• Python 3.8 or higher
• ffmpeg (for MP4 video processing)
• CUDA-compatible GPU (optional)

Installation

1. Clone or Download the Project

2. Install Python Dependencies

pip install -r requirements.txt

3. Install ffmpeg

Windows:

# Using chocolatey:
choco install ffmpeg

# Or download from: https://ffmpeg.org/download.html

macOS:

brew install ffmpeg

Linux:

sudo apt install ffmpeg

Usage

1. Prepare Your Files

• Place your files in the presentations folder:
  • PowerPoint presentations (.pptx)
  • Video files (.mp4)
  • Audio files (.mp3)

2. Run the Transcriber

python main.py

Configuration

Edit the configuration settings at the top of main.py:

Transcription Engine

TRANSCRIPTION_ENGINE = "faster-whisper"  # Options: "standard", "faster-whisper"

Folder Settings

PPTX_FOLDER = "presentations"   # Input folder
OUTPUT_FOLDER = "output"        # Output folder

Whisper Model Settings

WHISPER_MODEL = "small"       # Options: "tiny", "base", "small", "medium", "large"
FORCE_LANGUAGE = "en"         # Force language ("en", "es", "fr", etc.) or None

Performance Settings

FORCE_DEVICE = "cpu"          # Options: None (auto), "cpu", "cuda"
USE_HALF_PRECISION = False    # Enable fp16 for speed boost (GPU only)

Model Size Guide

Model	Speed	Quality	Memory	Best For
tiny	Fastest	Good	~1GB	Quick drafts, testing
base	Fast	Better	~1GB	General use
small	Medium	Good	~2GB	Recommended - best balance
medium	Slow	Very Good	~5GB	High accuracy needs
large	Slowest	Best	~10GB	Maximum quality