main.py

import openai
import os
import sys
from pydub import AudioSegment
import math
from quarter import split_audio
from summary import gptsummary, gpt_chat_completion_xl
import time
import logging

"""
This program processes an mp3 audio file and splits it into quarters. It requires FFMPEG. It then processes these chunks via OpenAi Whisper speech to text and creates an appended text file called transcript.txt and then it uses Openai GPT-3.5-turbo-16k to summarize the text into a few paragraphs when done. The user is required to fill in the starting mp3 audio file as the Audio_file_to_process variable below before running this program. Also be sure to configure your OpenAi API key in the Secrets before use. Too large of mp3 files may still max out RAM and produce an error unless system resources are boosted. Delete the transcript.txt file before use.
"""

####### Start here: ########
# USER MUST INPUT MP3 FILE NAME TO PROCESS INTO QUOTES BELOW:
Audio_file_to_process = "training-cb1.mp3"

# USER SHOULD EDIT THIS GUIDING PROMPT FOR DETERMINING SPELLING etc. for 
# THE WORDING OF THE TRANSCRIPTION
prompt_guide = "audio is the Iowa Ag Podcast with host, Peter Jaques"
###### End inputs.  Run program if ready #########


#logging config
logging.basicConfig(filename='app.log', filemode='a', format='%(asctime)s - %(name)s - %(levelname)s - %(message)s', datefmt= "%d-%b-%y %H:%M:%S", level=logging.INFO)
logger = logging.getLogger(__name__)

# make sure starting file size is less than 100 MB
file_size = os.path.getsize(Audio_file_to_process)
file_size_mb = file_size / (1024 * 1024)
if file_size_mb >= 100:
  print("audio file to process must be less than 100 MB")
  logger.error("audio file to process was greater than 100 MB. Too large")
  sys.exit(1)
  
# This imported function splits the beginning audio into 4ths
# first param is mp3 file to split into quarters and save
# into the segmented folder. 2nd param is folder to save to.
try:
  split_audio(Audio_file_to_process, "segmented") 
  time.sleep(10)

except:
  logger.error("issue with split_audio")

#OpenAi limit on uploaded mp3 file is 25 mb

try:
  openai.api_key = os.environ['OPENAI_API_KEY']
except KeyError:
  logger.error("Openai api key issue")
  sys.stderr.write("""
  You need to set up your API key yet.
  Manage API keys at:
  https://platform.openai.com/signup

  Open the Secrets Tool and add OPENAI_API_KEY as a secret.
  """)
  exit(1)

# Get a list of all files in the folder
file_list = os.listdir("segmented")

# delete old transcript.txt file if it exists
if os.path.exists("transcript.txt"):
  os.remove("transcript.txt")

# Create the transcript file from audio segments and append whisper results
with open("transcript.txt", "a") as transcript_file:
  for file_name in file_list:
    with open("segmented/" + file_name, "rb") as audio_file:
      try:
        transcript = openai.Audio.transcribe("whisper-1", audio_file, prompt=prompt_guide)
        print("The transcript of the audio file is:\n\n")
        print(transcript["text"])
        transcript_file.write(transcript["text"])
        print("\n\nTranscript saved to the file transcript.txt")
        time.sleep(10)
      except:
        logger.error("whisper transcription issue")

try:
  gptsummary("transcript.txt")
except:
  logger.error("problem running gptsummary()")