← All guides

Best free ways to get a YouTube transcript

Written by the VidWords Team · June 11, 2026 · Updated October 8, 2026 · Report a correction

Every free route, the named tools people actually use, and the cases where you need no service at all. We make one of these tools; the comparison below is built so you can check it rather than trust it.

The short answer

• One video, read it once — YouTube's own …more → Show transcript panel. Free, official, no export.
• One video, want the file — any free transcript site. Most take a URL and hand back .txt without an account.
• A video with no captions, or bad ones — a speech-recognition service, which is a different kind of tool entirely. See the two architectures below.
• Writing code — youtube-transcript-api or yt-dlp, moving to a hosted API when YouTube starts blocking your server's IP.

The distinction that decides everything

Tools in this category look interchangeable and are not. They split into two groups that fail in opposite ways.

Caption extractors read the caption track YouTube already stores. That is a file fetch: it is near-instant, costs almost nothing, works identically on a three-minute clip and a three-hour lecture, and carries the video's chapter markers. It is also completely helpless when no caption track exists.

Speech-recognition services ignore the captions and run a model over the audio. That works on any video, usually beats YouTube's automatic captions on punctuation and proper nouns, and can label speakers. It also costs compute, which is why these tools price and cap by duration — and why their free tiers tend to cap the length of a single file.

Most "best transcript tool" lists mix the two together and then compare their free tiers, which is how you end up with a recommendation that cannot transcribe your two-hour conference talk.

1. YouTube's built-in transcript viewer

Under any video: …more → Show transcript. Free and official, and the only method involving no third party at all. On desktop, open a video, expand the description, and click Show transcript: a panel appears beside the player listing the captions line by line, and clicking a line jumps the video to that moment.

What it has no button for is getting the text out. Your only route is to drag-select the panel, paste it somewhere, then strip hundreds of timestamps and rejoin the broken lines by hand. There is no export, no search across videos, and nothing for playlists or channels. The panel exists on mobile too, where selecting text out of it is fiddlier still.

2. Open-source libraries

If you write code, youtube-transcript-api (Python) and yt-dlp are excellent, free, and have no quota at all. You handle formatting, retries, and — at any real volume — the thing everyone hits eventually: YouTube blocks datacenter IPs, so code that works perfectly on your laptop fails the moment it runs on a server. That is not a bug in the library, and no amount of retrying fixes it. We keep separate write-ups on why the block happens and what actually clears it, and on yt-dlp's "confirm you're not a bot" wall.

yt-dlp can also save raw .vtt/.srv caption files directly. Accurate, but raw: duplicated rolling-caption lines and cue formatting to clean up.

3. The official YouTube API — the route that doesn't exist

It is the first thing most developers try, so it is worth ruling out explicitly. The YouTube Data API v3 has a captions resource, but the official captions.download documentation states the method requires OAuth authorization and permission to edit the video — it exists so authorized creators can manage their own caption tracks. captions.list will tell you which tracks exist and in which languages; it will not hand over the words. For a public video you do not control there is no API-key-only official endpoint, which is why every other method here is, in one form or another, a workaround.

4. The hosted tools, compared

These are the products that come up most often, including in AI-assistant recommendations. Everything was re-checked on 8 October 2026 against each tool's own pages. The exceptions: DownSub and Kome serve no readable text to an automated check, so their rows carry the 26 September reading, and TurboScribe refuses automated requests, so its row comes from published trackers. Prices and limits move, so verify anything you are about to depend on. Blank cells mean the tool does not publish a figure — which is itself worth knowing.

Source ledger: YouTube transcript help; Tactiq YouTube tool; NoteGPT pricing; TurboScribe pricing; Descript pricing; DownSub; and Kome transcript tool; youtube-transcript.io pricing; NoteGPT summariser. The "Best for" column is our judgement, not the vendor's claim. “Not documented” means no public documentation was found, not that a logged-in feature cannot exist. Verify current terms before choosing a paid service.

Two things that table makes obvious. First, almost everything free exports .txt and nothing else — fine if a human is going to read it, useless if the text has to go back onto a video as subtitles or into a script as structured data. Second, batch is where the field thins out: most of these tools are built around one URL at a time, so a channel is a hundred separate visits.

Where a tool is genuinely the better pick, we have written that down: Tactiq if your real problem is meeting transcription, TurboScribe when a video has no usable captions, Descript when you are editing the video, and NoteGPT when you want one suite covering PDFs and slides too.

Getting from raw captions to usable text

Whichever free route you take, the same problem waits at the end. YouTube stores captions as short timed fragments of one to five seconds each, sized for display along the bottom of a player rather than for reading, so dumping them out gives you no sentences, no paragraphs, and a timestamp every few words. Automatic tracks add their own quirks: sparse punctuation, no speaker labels, and duplicated lines wherever rolling captions scroll. Budget for that cleanup, or pick a method that does the merging for you.

Which export format you want depends on where the text is going. TXT is just the words, for reading, quoting, or pasting into an LLM. SRT is the universal subtitle format that video editors and media players accept. VTT is the web-native equivalent that HTML5 <track> elements expect. CSV puts one cue per row with its timings, so a spreadsheet can sort and annotate it. JSON is the same cue data structured for scripts and search indexes. A rough rule: SRT or VTT when the file goes back next to a video, TXT when a human will read it, CSV or JSON when software will — and a method that produces only one of them eventually costs you a conversion step. Our guide to downloading YouTube subtitles as SRT, VTT or TXT goes format by format.

Which should you use?

When none of them will work

No caption-based method on this page can retrieve a track that does not exist. That covers a video uploaded so recently that captions are still processing, a music video with little recognizable speech, a video whose creator supplied no captions and for which no automatic track was produced, speech in a language YouTube does not auto-caption, and anything age-restricted, private, or members-only.

For the first four, speech recognition over the audio is the remaining option. For the last, nothing legitimate will help — the video is not public. Auto-captions explained covers how to tell which case you have hit, and transcript not showing works through the symptoms one at a time.

Get a free transcript →