Recipe: YouTube video summary

How it works

1. youtube-dl downloads the audio track as a WAV file.
2. whisper transcribes the audio to text using the base model.
3. The transcript is passed to an LLM with a summarization prompt.
4. The summary is returned as structured Markdown.

Prerequisites

• Python 3.10+
• yt-dlp
• openai-whisper
• OpenAI API key (or local LLM)

Pipeline script

#!/usr/bin/env python3
import subprocess, whisper, openai

URL = "https://www.youtube.com/watch?v=VIDEO_ID"
AUDIO = "audio.wav"

subprocess.run(["yt-dlp", "-x", "--audio-format", "wav",
                "-o", AUDIO, URL], check=True)

model = whisper.load_model("base")
result = model.transcribe(AUDIO)
transcript = result["text"]

response = openai.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{
        "role": "system",
        "content": "Summarize this transcript concisely."
    }, {
        "role": "user",
        "content": transcript
    }]
)

print(response.choices[0].message.content)

Usage

python summarize.py

Replace VIDEO_ID with the target YouTube video ID. Output prints directly to stdout.