an example on how to save an hour of one's time using openai's speech 2 text & then gpt4-o to summarize and ask questions about the resulting text

chunk_audio.py

#!/usr/bin/env python
from pydub import AudioSegment
from pydub.silence import split_on_silence
import sys
import os

def chunk_audio(input_file, output_prefix, min_silence_len=len(sys.argv)>2 and int(sys.argv[2]) or 1000, silence_thresh=-40, keep_silence=None):
    if not keep_silence:
        keep_silence = min_silence_len/4
        
    audio = AudioSegment.from_mp3(input_file)

    # Split audio where silence is 'silent' for at least min_silence_len ms and lower than silence_thresh dBFS
    chunks = split_on_silence(
        audio,
        min_silence_len=min_silence_len,
        silence_thresh=silence_thresh,
        keep_silence=keep_silence
    )

    # Export chunks as separate files
    for i, chunk in enumerate(chunks):
        chunk_name = f"{output_prefix}{str(i).zfill(3)}.mp3"
        chunk.export(chunk_name, format="mp3")
        print(f"Exported {chunk_name}")

if __name__ == "__main__":
    if len(sys.argv) < 2:
        print("Usage: python chunk_audio.py <input_file> [min_silence_len]")
        sys.exit(1)

    input_file = sys.argv[1]
    output_prefix = "output_chunk_"
    chunk_audio(input_file, output_prefix)

run.sh

#!/bin/bash
yt-dlp --extract-audio 'https://www.youtube.com/watch?v=eGPa_omV9WI'
ffmpeg -i *m4a -c:v copy -c:a libmp3lame -q:a 4 input.mp3
chunk_audio.py input.mp3
for fn in output_chunk*mp3 ; do echo " * $fn" ;  transcribe.py "$fn" | tee "$fn.txt" ; done
cat output_chunk*txt > output.txt
aider output.txt summary.txt -m 'summarize this into `summary.txt`'

transcribe.py

#!/usr/bin/env python
from openai import OpenAI
import sys
import os

# Initialize the OpenAI client
client = OpenAI(api_key=os.environ.get("OPENAI_API_KEY"))

# Function to transcribe audio using OpenAI's latest API
def transcribe_audio(audio_file_path):
    with open(audio_file_path, 'rb') as audio_file:
        response = client.audio.transcriptions.create(
            model="whisper-1",
            file=audio_file
        )
    return response.text

# Example usage
if __name__ == "__main__":
    if len(sys.argv) < 2:
        print("Usage: ./transcribe.py <audio_file_path>")
        sys.exit(1)
    
    audio_file_path = sys.argv[1]
    transcription = transcribe_audio(audio_file_path)
    print(transcription)

results in:
summary.txt

Chris Hedges, a Pulitzer Prize-winning journalist, discusses the current state of American politics, focusing on the corporate influence over both major parties. He critiques the Democratic Party's handling of
Biden and Harris, highlighting their ties to corporate interests and the lack of genuine voter influence. Hedges also addresses the rise of Trump, attributing it to widespread despair and betrayal by the
political system. He emphasizes the systemic issues within the U.S., including economic decay, militarism, and the influence of the Israel lobby. Hedges calls for a movement to challenge the political hegemony
from outside the established parties, advocating for a focus on addressing the needs of the vulnerable and ending corporate exploitation.