1.3k stars · MIT · 0.1.83 (2026-07-30)
Every comment under any YouTube video as JSON — no key, no quota. The complaints under competitors' tutorials are your next ten video titles.
▶Repo detailsthe review · specs · pros & cons · install
What it is
A small Python tool that fetches the comments from any YouTube video — text, author, timestamp, like count, and replies — as JSON lines, sorted by popularity or recency. No API key, no quota, no OAuth consent screen; it is a command you run and a file you get.
Why it matters
This is the sharpest research tool this radar has found, and it is nearly free. Edition 6 gave you yt-dlp for what competitors published and Edition 5 gave you youtube-transcript-api for what they said. Comments are the other side of that conversation, and they are the only one of the three written by your buyers. Pull the comments from the twenty most-viewed videos about summarising long chats, mapping documents and backtesting forex strategies, and you will find the same six complaints repeated in different words — "he goes too fast", "this doesn't work on Mac", "what about a 500-page PDF", "how do I do this in Russian". Each of those is a video title, a docs page, and a line of landing-page copy, and each one is evidence rather than instinct. It also works on your own channel: the questions under your videos are the fastest signal you have about what the product fails to explain — and they are also the thumbnail and title feedback loop, because a comment saying "clicked for X, got Y" is telling you your title lied. Note the trap this project illustrates: its GitHub releases page stops at August 2025 and looks abandoned, while PyPI shows a release five weeks ago. Check PyPI for this one.
- No key, no quota, no consent screen — and it burns none of the 10,000-unit daily Data API budget that constrains python-youtube
- MIT, tiny, pure Python, and it emits JSON lines that go straight into a file, PostgreSQL, or the model in entry 07
- Comments are the one competitor signal that comes from the audience rather than the marketer — this is the closest thing to a free focus group you will ever run
- Unofficial endpoint, exactly like youtube-transcript-api. YouTube can change it, and datacentre IP ranges get blocked — run it from your Mac, not the Hetzner box
- Seventy-eight commits and one maintainer. It is small enough that it either works or is replaceable in an afternoon, but do not build a product feature on it
- Comment sections are noisy: spam, jokes, and "first". Expect to throw away most of what you pull, which is the job entry 07 exists to do for you
Install into a virtual environment on the Mac beside your Edition 6 research tools.
source venv/bin/activate pip install youtube-comment-downloader # the 500 most-liked comments on a competitor's tutorial youtube-comment-downloader --url "https://www.youtube.com/watch?v=VIDEO_ID" \ --sort 0 --limit 500 --output comments.json