Youtube Script Summarizer

Overview

Youtube Script Summarizer is a tool that helps summarize YouTube scripts by chapter creators. It leverages the youtube-transcript-api, langchain, and OpenAI packages to achieve this.

Dependency

• Python 3.11.x
• yt-dlp
• pydub
• OpenAI-Whisper
• langchain
• OpenAI

How it works

• Obtain the audio file by using yt-dlp
• Divide the audio file into chapters with corresponding information by using pydub
• Perform transcription on each audio segment utilizing OpenAI-Whisper
• Summarize the transcriptions using the langchain and OpenAI packages.

Usage

0. Clone this project and setup Python over v3.11.x.

1. Open setup.ipynb and run to setup project.

2. Setup OPENAI_API_KEY.

• Create .env file and set OPENAI_API_KEY variable.

3. Open transcript.ipynb

• Create .env file and set OPENAI_API_KEY variable.
• Edit the variables in the "Input variables" cell according to your requirements.
  • Provide the YouTube video ID (youtube_video_id) for which you want to generate a summary.
  • Specify the language of the subscription (language).
  • Copy and paste the time stamps and chapter names from the YouTube video description into the chapter_part_in_description variable.

# Youtube video ID
youtube_video_id="MZQ6bc6mPAE"

# Language of subscription 
language = "ko"


# Officially no way to get chapter automatically, 
# so copy and paste the time stamp and chapter in description of Youtube video. 
chapter_part_in_description = """
00:00 시작
05:24 해먹을 결심: 탄핵해야 하는 이유
10:55 댓글 읽어보기
24:27 인사조직론이란 무엇인가?
33:52 (게르만 모형) 왜 직무가 중요한가?
43:11 추미애의 직무인식
49:16 직무의 존재목적, 칸트의 인간관과 경영학적 응용
1:04:40 성과책임의 사회적 의미 (참고)직무의 3대 구성요소
1:06:05 직무개념의 부재
1:07:44 주진우와 양향자
1:11:52 추미애의 고백과 진심
1:35:09 왜 역량인가?
1:35:47 역량의 개념에 대한 이해
1:38:05 《성취예측모형》 프레임워크와 역량사전
1:41:48 진실한 리더십과 인재평가의 프레임워크
1:43:59 DANO 경영플랫폼 운용_리더십이란 무엇인가?
1:49:42 추미애에 대한 오해의 프레임과 추미애의 비전은 무엇인가?
1:53:21 이재명과 추미애 vs. 이낙연과 김진표
1:55:52 푸른 하늘을(김수영 시인, 1960.06.15.)
1:58:57 정리
"""

3. Run summarize_note.ipynb notebook and you can get the markdown_note.md.

   • This file has the script and summary for each chapter.

4. Run clean_data.ipynb to clean up the temporary and result data.

Trobleshooting

• Please define '00:00:00' as start in chapter_part_in_description.
• Keep the duration of chapter within 20 minutes for stable performance.
  • If some chapter is longer than 20 minutes, then add more chapters temporary.
• Need

FAQ

• To-Do