Paste a video id or a watch URL and get the comment section back as a table - who wrote it, their channel id, the text, the likes, the reply count, and whether the row is a reply or a top-level comment. Sort by Top or Newest, and pull replies too if you want them.
There is no page to load and no cookie to supply. You give it a video, and choose how much of the comment section you want.
The service reads YouTube's own JSON interface rather than parsing the page HTML, so a layout change on YouTube does not silently break the export.
Top-level comments come back on their own by default. Tick Extract replies and the reply rows arrive too, each one flagged in is_reply so you can separate the conversation from the thread openers without guessing.
The same video sorted by Top and by Newest answers two different questions - what the audience upvoted, and what they are saying now. The sort is a parameter, so you can pull both and compare them.
The limit applies per query, so one job can take the first hundred comments from each of many videos. Set it to 0 and it takes the comment section as far as it goes.
The names below are the header row of the export, exactly as it arrives. One row per comment. The descriptions say what each column is for and how it actually behaves - including the two places where the export will surprise you.
query - the same video submitted two different ways produces the same video_id.yes or no, not a boolean - so test for the string.The JSON export carries one field the spreadsheet does not: position. It is present in the header of every .json we have and absent from every .xlsx. If you are building against this service and want the ordering the comment section came back in, take the JSON - converting the spreadsheet will not give it to you.
Where a comment section is more useful as a spreadsheet than as an infinite scroll.
A comment section is unprompted, unmoderated feedback on a specific thing at a specific time. Pull it for your own videos and for the ones your audience watches instead, sort by likes, and the questions and objections people actually have rise to the top on their own.
Take the comments with a high reply_count and no reply from the channel - those are the threads the audience is already discussing without you. It is a short list, and working it is the cheapest engagement there is.
Run the same video twice, once sorted by Top and once by Newest. The gap between what an audience upvoted and what it is saying now is where sentiment has moved - and it is invisible if you only ever look at one ordering.
No subscription, no minimum, no per-seat licence. Your first 500 rows are on us - after that it is pay-as-you-go.
For every new account, one time. No credit card. All scrapers unlocked. One row is one comment, so the free tier covers a real comment section rather than a token sample.
Roughly $2 per 1,000 comments. Because the limit is per query, the cost of a job is something you set rather than something you discover - cap it while you are exploring, lift it once you know the video is worth taking in full.
Volume pricing, SLAs, dedicated workers and tailored onboarding for teams tracking comments across many videos on a repeating schedule. Tell us your numbers and we will quote.
LIVESCRAPER10 at checkout.What was said underneath, what was said in it, and who published it - three different questions.
The videos themselves - from a channel, a playlist or a keyword search. Use it to build the list of video ids you then feed to this scraper.
The full spoken text of a video. Comments tell you how an audience reacted; the transcript tells you what they were reacting to.
Channel-level data - subscribers, views and video count, already parsed to numbers. The context behind whoever published the video.
Short answer: yes, when you only collect what is publicly visible under the video - and that is all this service reads.
A comment section is public. Anyone can read it without an account, and collecting publicly visible information for research is long-settled practice. As long as the data is publicly available and the process does not disrupt the service, there are no federal laws against it.
Be aware of what this data is, though, because it is not like a price or a listing. A comment is written by a person, and author and author_channel identify them. Publicly visible is not the same as free of obligation - if you store it, you are handling personal data, and the GDPR and similar regimes apply to you regardless of where you got it. Collect what you need, keep it as long as you need it, and do not republish individuals' comments alongside their channel ids without thinking about it first.
YouTube's terms restrict automated access, so this remains a question of terms. We touch nothing behind a login, read only what an ordinary viewer sees, run no third-party trackers on the data layer, and your exports self-delete after 30 days.
The questions we hear most. Something else? Talk to us - humans write the answers, not bots.
Hand it a video id or a watch URL and take the comments away as rows - author, channel id, text, likes and reply count. Your first 500 rows are free.
A YouTube comment section is one of the few places where an audience says what it thinks without being asked. It is also built for scrolling rather than reading: there is no way to sort it your way, no way to filter it, and no way to compare one video against another. This service turns it into a table - one row per comment, with the author, their channel id, the comment text, the like count, the reply count and a flag for whether the row is a reply, exported as CSV, JSON or Excel.
Driving it is deliberately plain. Paste video ids or full watch URLs, one per line, or upload a CSV, TXT or XLSX and the first column is read as your list. A limit applies per query, so a single job can take the first hundred comments from each of many videos, or take a whole comment section when you set it to zero. The sort is yours - Top or Newest - and replies are a separate toggle that is off unless you ask for them. The service reads YouTube's own JSON interface rather than parsing page HTML, so a layout change on the site does not quietly break your export.
Two things about the data are worth knowing before you build on it. The published column is not a date: it carries the relative text YouTube itself displays, at whatever unit fits how old the comment is - "1 day ago" on a fresh comment, "2 years ago" on an old one - with an "(edited)" suffix where it was edited, so it will not parse as a timestamp and is only meaningful next to the date you ran the job. And likes and reply_count arrive blank rather than zero when nothing is shown, which will quietly distort a sum or an average unless you coalesce first. The is_reply column is the string yes or no rather than a boolean, so test for the string.
One honest note about the schema. The spreadsheet export has ten columns and the JSON has eleven - the JSON adds position, the ordering the comment section came back in, and converting the spreadsheet will not recover it. And parent_author, which is meant to name the author a reply is answering, came back empty on every row of every run we made, including the reply rows where it is the only field that could be populated. It is documented because it is part of the declared schema, not because we have seen it work; check your own export before you depend on it. Start free: your first 500 rows cost nothing and need no credit card. See pricing for current rates.