Audio Transcription with Whisper
This guide demonstrates how to use OpenAI's Whisper, a robust open-source speech recognition model. It allows you to transcribe audio files into text locally on your machine, supporting multiple languages and accents.
Modules Used:
openai-whisper: The Python library for the Whisper model.- argparse: To handle command-line arguments.
os: For file path handling.
Prerequisites
You must have FFmpeg installed on your system.
- Windows:
winget install ffmpeg - macOS:
brew install ffmpeg - Linux:
sudo apt install ffmpeg
Installation
pip install openai-whisperThe Code
Save this as transcriber.py.
import whisper
import argparse
import os
import warnings
# Suppress FP16 warning on CPU-only machines
warnings.filterwarnings("ignore")
def transcribe(audio_path, model_size, output_file):
if not os.path.exists(audio_path):
print(f"Error: File '{audio_path}' not found.")
return
print(f"Loading Whisper model ('{model_size}')...")
try:
# Models: tiny, base, small, medium, large
# The first run will download the model weights
model = whisper.load_model(model_size)
except Exception as e:
print(f"Error loading model: {e}")
return
print(f"Transcribing '{audio_path}'... (This may take time)")
try:
# Transcribe
result = model.transcribe(audio_path)
text = result["text"]
# Output
if output_file:
with open(output_file, "w", encoding="utf-8") as f:
f.write(text.strip())
print(f"Success! Saved to '{output_file}'")
else:
print("\n--- Transcription ---")
print(text.strip())
print("---------------------")
except Exception as e:
print(f"Error during transcription: {e}")
if __name__ == "__main__":
parser = argparse.ArgumentParser(description="Whisper Audio Transcriber")
parser.add_argument("audio", help="Path to the audio file (mp3, wav, m4a, etc.)")
parser.add_argument("--model", default="base", choices=["tiny", "base", "small", "medium", "large"], help="Model size (default: base)")
parser.add_argument("-o", "--output", help="Output text filename")
args = parser.parse_args()
transcribe(args.audio, args.model, args.output)Usage
# Transcribe and print to console
python transcriber.py meeting.mp3
# Use a larger model and save to file
python transcriber.py interview.wav --model medium -o interview.txt