Use ffmpeg, Whisper and other OpenAI LLM to transcribe and summarize a video

README.MD

Video Transcriber

This project extracts audio from a video file and transcribes it using the OpenAI Whisper API, and uses a prompt with another model of your choice

Prerequisites

• Python 3.10+ (tested with Python 3.13)
• pip (Python package installer)
• An OpenAI API Key with access to audio transcription

Setup

1. Create a Virtual Environment

   It’s recommended to use a virtual environment:

   python3 -m venv env
   source env/bin/activate
   

2. Install Dependencies

   pip install requests
   pip install openai
   

3. Set Up Environment Variables

   The script requires an OpenAI API key for transcription. Set the OPENAI_API_KEY environment variable:

   • On macOS/Linux:

     export OPENAI_API_KEY=your_openai_api_key
     

   • Alternatively, create a .env file in the project root containing:

     OPENAI_API_KEY=your_openai_api_key
     

     and use a package like python-dotenv to load it in your script.

Usage

1. Extract audio and transcribe:

   Run the transcriber script:

   python transcriber.py INPUT_FILE --prompt "Please summarize this"
   

2. Process Overview

   • The script extracts the audio from the video and saves it as temp_audio.mp3.
   • It sends the extracted audio file to the OpenAI Whisper API for transcription.
   • It takes the output of the transcription into another LLM Prompt of your choice, using gpt-4o-mini as the default model
   • If the API request fails with a 401 error, check your API key and ensure it is set correctly.

video_transcribe.py

#!/usr/bin/env python3

import os
import subprocess
import argparse
import openai
import requests

openai.api_key = os.getenv("OPENAI_API_KEY")

def extract_audio_from_video(video_path, audio_path):
    command = [
        "ffmpeg", "-i", video_path,
        "-vn",  # No video
        "-acodec", "mp3",
        audio_path,
        "-y"
    ]
    subprocess.run(command, check=True)

def transcribe_audio_whisper(audio_path):
    """
    Transcribe an audio file using OpenAI's Whisper model.
    Uses direct HTTP request call to OpenAI's transcription endpoint.
    """
    url = "https://api.openai.com/v1/audio/transcriptions"
    headers = {
        "Authorization": f"Bearer {openai.api_key}"
    }
    data = {
        "model": "whisper-1"
    }
    with open(audio_path, "rb") as audio_file:
        print("[INFO] Transcribing audio with Whisper...")
        files = {
            "file": audio_file
        }
        response = requests.post(url, headers=headers, data=data, files=files)
    response.raise_for_status()
    result = response.json()
    return result["text"]

def summarize_transcript(transcript, user_prompt):
    """
    Summarize the transcript via OpenAI's Chat Completion API using a direct HTTP request.
    """
    url = "https://api.openai.com/v1/chat/completions"
    headers = {
        "Authorization": f"Bearer {openai.api_key}",
        "Content-Type": "application/json"
    }
    system_content = "You are a helpful assistant..."
    user_content = (
        f"The transcript of the video is:\n\n{transcript}\n\n"
        f"User prompt: {user_prompt}\n\n"
        "Please summarize or analyze the transcript."
    )
    payload = {
        "model": "gpt-4o-mini",
        "messages": [
            {"role": "system", "content": system_content},
            {"role": "user", "content": user_content},
        ],
        "temperature": 0.7
    }
    response = requests.post(url, headers=headers, json=payload)
    response.raise_for_status()
    result = response.json()
    return result["choices"][0]["message"]["content"]

def main():
    parser = argparse.ArgumentParser(description="Video Summarizer with OpenAI Whisper + GPT")
    parser.add_argument("video_file", help="Path to the local video file")
    parser.add_argument("--prompt", type=str, default="Please summarize...",
                        help="User-defined prompt to guide the summary.")
    args = parser.parse_args()

    video_file = args.video_file
    user_prompt = args.prompt

    print(f"[INFO] Extracting audio from {video_file}...")
    audio_file = "temp_audio.mp3"
    extract_audio_from_video(video_file, audio_file)
    print(f"[INFO] Audio extracted to {audio_file}")

    transcription_text = transcribe_audio_whisper(audio_file)
    print("[INFO] Transcription complete.")

    summary = summarize_transcript(transcription_text, user_prompt)
    print("[INFO] Summary/Analysis complete.\n")

    print("TRANSCRIPTION:\n", transcription_text)
    print("\n----------------------------------------\n")
    print("SUMMARY / ANALYSIS:\n", summary)

if __name__ == "__main__":
    main()