Skip to main content
NoteWise logo NoteWise is a command-line tool that turns YouTube content into comprehensive Markdown study notes — powered by the LLM of your choice, with no browser required.
Built for learners and builders who want repeatable, automatable study workflows from long-form video content.

​
Jump in

Quick Start

Get your first AI-powered study notes in under 5 minutes.

Installation

Install via pip, uv, or Docker.

CLI Reference

Every command, option, and flag documented.

Configuration

All settings, defaults, and override mechanisms.

​
Choose your workflow

notewise process "https://youtube.com/watch?v=VIDEO_ID"
Best for focused notes from one lecture or tutorial.

​
How it works

1

Fetch the transcript

Retrieves the video transcript with automatic language fallback and retries. Supports private and age-gated videos via cookie file.
2

Detect chapters

Long videos (over 1 hour with chapter metadata) are split into chapters and processed concurrently for better quality.
3

Generate notes

The transcript is chunked intelligently and sent through your configured LLM via LiteLLM. Supports Gemini, OpenAI, Anthropic, Groq, and more.
4

Write structured Markdown

Study notes are written as Markdown files named from the video title. Optionally generates quizzes and transcript exports.
5

Cache locally

Transcripts and run stats are stored in a local SQLite database. Already-processed videos are skipped on re-runs automatically.

​
Key features

Multi-provider LLM

Gemini, OpenAI, Anthropic, Groq, Mistral, Cohere, DeepSeek, and xAI — all via LiteLLM. Switch models with a single flag.

Concurrent processing

Multiple videos and chapters processed in parallel. Configurable concurrency keeps you within API and YouTube rate limits.

Local SQLite cache

Transcripts and run stats persist across runs. Re-run a batch safely — only unprocessed videos are touched.

Rich terminal UI

Live progress dashboard. Use --no-ui for CI, cron, or log aggregators.

Playlist & batch support

Process entire playlists or a .txt file of URLs. Re-runs skip already-processed videos unless --force is used.

Docker-ready

Two-stage image published to GHCR on every release. Runs as a non-root user.
The default model is Gemini 2.5 Flash — free tier available via Google AI Studio. No billing required to get started.

​
Example session

$ notewise process "https://youtube.com/watch?v=dQw4w9WgXcQ"

 notewise  v1.0.0

Processing 1 video...

 Video Title
   ✔ Metadata fetched
   ✔ Transcript fetched  (en, 342 segments)
   ✔ Notes generated  (3 chunks)
   ✔ Written → output/Video Title.md

────────────────────────────────────────
  1 processed · 0 failed · 12.4s · $0.0004

​
Explore by role

New user

Start with Quick Start → Installation → Configuration.

Power user

Jump straight to the process command reference.

DevOps / CI

Docker and --no-ui patterns for automation.

Contributor

Dev setup, testing, and CI/CD docs.