LangChain Video Transcription & Study Material Generator

Overview

This project began as a simple experiment: transcribing and understanding a few educational videos.
It quickly evolved into a fully automated pipeline that transforms raw video content into rich learning material — complete with transcripts, structured summaries, glossaries, and practice questions — all orchestrated seamlessly.

The result is a system that demonstrates not only technical execution but also the architectural thinking required to design scalable, reliable, and extensible solutions.

---

Features

• Video → Audio Extraction: Uses ffmpeg to isolate audio streams from video files.
• Audio → Transcript: Employs OpenAI Whisper for accurate speech-to-text transcription.
• Transcript → Summary: Leverages LangChain with Ollama LLM models to generate structured summaries with topics, headings, and concepts.
• Summary → Study Material:
  • Key concepts and definitions
  • Bullet-point notes
  • Glossary of important terms
  • 25 practice questions (MCQ + short answer)
• Markdown → PDF: Converts study material into polished PDFs using Pandoc + LaTeX.
• CLI Experience: Clean, readable, single-line progress updates with spinners for in-progress stages.

---

Tools & Technologies

Stage	Tool/Technology	Purpose
Video → Audio	ffmpeg	Reliable audio extraction
Audio → Text	Whisper	State-of-the-art transcription
Text → Summary	LangChain Core + Ollama LLM	Prompt orchestration & summarization
Summary → Study Material	LangChain PromptTemplate	Rich, structured learning content
Markdown → PDF	Pandoc + LaTeX	Professional document generation
CLI UX	Python sys.stdout + spinner	User-friendly progress visualization

---

Orchestration

The pipeline is designed as a modular sequence of transformations:

video → audio → transcript → summary → study material → PDF

Each stage:

• Checks for existing artifacts (skip if already generated).
• Runs independently but feeds into the next stage.
• Updates CLI output in place, showing progress with spinners.
• Produces reusable artifacts (.txt, .md, .pdf) for downstream use.

This orchestration ensures idempotency, scalability, and clarity — critical traits for production-ready automation.

---

Why This Project Showcases Engineering Excellence

• Deep Research: Started with exploring Whisper and LangChain, evolved into a full pipeline.
• Solution Design: Identified pain points (manual transcription, scattered notes) and designed a structured workflow.
• Architecture: Modular, fault-tolerant, idempotent design with clear separation of concerns.
• Automation for Scale: One command processes entire directories of videos, producing consistent, high-quality study material.
• User Experience: Thoughtful CLI design with progressive spinners and clean output.
• Compliance & Reliability: Uses open-source, well-supported tools with reproducible environments (requirements.txt).

---

Example CLI Output

Each stage shows a rotating line as progress bar and ellipsis to denote in-progress

AI is warming up... ready to crunch some knowledge.

myVideo.mp4
    video > ... |
    video > audio > ... /
    video > audio > text > markdown > study > PDF

---

Installation

python3.11 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Usage

python langchain_videos_transcription.py ./videos --pdf