A very basic processing of audio files text to speech using Whisper AI in Python (running on your computer)

basic-pipe-whisper.py

## Converts Speech to Text using the OpenAI Whisper (English medium size ~3GB) AI model
## NOTE: You'll want to update the file (e.g "TheUnaccountabilityMachine.mp3") to whatever mp3, flac or wav file that you want processed.

## -- Instal requirements by installing Python 3 and pip then run:
# pip install torch transformers librosa soundfile ffmpeg

## -- Run in Windows (after updating the filenames in the code below) using:
# cd C:\workspace\whisper-huggingface\
# python basic-pipe-whisper.py

import os,sys
import subprocess
import glob
import json
from os import path
from datetime import datetime
from transformers import pipeline

now = datetime.now()
print("The Starting Time is: ", now)

pipe = pipeline("automatic-speech-recognition", "openai/whisper-large-v3")
result = pipe("TheUnaccountabilityMachine.mp3", return_timestamps=True)

# NB: We have return_timestamps=True to support audio longer than 30s, as it's then split into chunks via the Hugging Face pipeline

print ("-------------------")
print(result)
print ("-------------------")

f = open('TheUnaccountabilityMachine.json','w')
f.write(json.dumps(result))

# Save just the plain text version, without the 
f = open('TheUnaccountabilityMachine.txt','w')
f.write(json.dumps(result['text']))

print("Converted and Saved to file")
print("Finished at ", datetime.now())