한국어 |

비디오-to-notes

video-to-notes는 전문적인 중국 LaTeX 코스 노트와 컴파일 된 PDF로 강의 소스 자료를 도는 Codex 기술 및 헬퍼 툴킷입니다.

주요 작업 흐름은:

YouTube URL과 cookies.txt이 제공된 경우, 워크플로우는 TXT로 변환하고 로컬 슬라이드로 transcript를 사용합니다.

저장소 레이아웃

• SKILL.md: Codex 기술 지침. 이것은 쓰기, 그림 선택, 자막 다운로드, LaTeX 및 검증 워크플로를 위한 진실의 근원입니다.
• assets/notes-template.tex: 간단한 제목 페이지, 하이라이트 상자, 목록, 그림 및 TikZ를 포함하여 기본 중국어 LaTeX 메모 템플릿.
• scripts/prepare_slide_images.py: 슬라이드 PDF를 pic/page_N.png 이미지로 렌더링합니다. 현재 디렉토리에서 하나의 PDF 및 하나의 TXT를 자동 발견 할 수 있습니다.
• scripts/srt_to_txt.py: .srt] 또는 .vtt] 자막을 읽을 수 있는 .txt 성적표로 변환합니다.
• clean.sh: 메모가 생성된 후 지정된 폴더로 현재 강의 작업공간을 아카이브합니다.
• agents/openai.yaml: 기술에 부합하는 짧은 에이전트-패싱 항목 포인트.

English | 简体中文

video-to-notes

video-to-notes is a Codex skill and helper toolkit for turning lecture source material into professional Chinese LaTeX course notes and a compiled PDF.

The main workflow is:

transcript TXT + slide PDF/PPT/PPTX -> slide images -> LaTeX notes -> rendered PDF

If a YouTube URL and cookies.txt are provided, the workflow can first download subtitles only, convert them to TXT, and then use the transcript with the local slides.

Repository Layout

• SKILL.md: the Codex skill instructions. This is the source of truth for the writing, figure-selection, subtitle-download, LaTeX, and validation workflow.
• assets/notes-template.tex: the default Chinese LaTeX note template, including a simple title page, highlight boxes, listings, figures, and TikZ.
• scripts/prepare_slide_images.py: renders a slide PDF into pic/page_N.png images. It can auto-discover one PDF and one TXT in the current directory.
• scripts/srt_to_txt.py: converts .srt or .vtt subtitles into a readable .txt transcript.
• clean.sh: archives the current lecture workspace into a named folder after notes are generated.
• agents/openai.yaml: a short agent-facing entry point for invoking the skill.
• demo/: sample slide PDFs for testing the image-preparation helper.

Supported Inputs

• Local transcript TXT plus local slide PDF.
• Local transcript TXT plus local slide PPT/PPTX, after converting the deck to PDF.
• YouTube URL or url.txt plus cookies.txt, then downloaded subtitles plus local slides.

When there are multiple candidate transcripts or slide decks, choose the files explicitly instead of relying on auto-discovery.

Prerequisites

Python helpers use only the standard library except for optional PDF rendering support:

python3 -m pip install pdf2image

Install Poppler so PDF pages can be rendered. On macOS:

brew install poppler

On Ubuntu/Debian:

sudo apt-get install poppler-utils

For optional YouTube subtitle download, install yt-dlp:

python3 -m pip install yt-dlp

For final PDF compilation, install a LaTeX distribution with XeLaTeX and common packages such as ctex, tcolorbox, listings, tikz, and pgfplots.

Quick Start

1. Prepare Source Files

Place the lecture transcript and slide deck in the working directory. If the transcript is in subtitle format, convert it first:

python3 scripts/srt_to_txt.py lecture.srt transcript.txt

2. Render Slide Images

If the working directory contains exactly one .pdf and one .txt, use auto-discovery:

python3 scripts/prepare_slide_images.py --auto

Otherwise, pass the slide PDF explicitly:

python3 scripts/prepare_slide_images.py --pdf slides.pdf --output pic