How to Save YouTube Captions to a Text File (.txt)

“Save YouTube captions to a text file” sounds like one step and is really three: get the track out of the video, get rid of the subtitle machinery around it, and make the resulting .txt open correctly on every machine you and your colleagues own. The first part is trivial. The other two are where people lose an afternoon — usually over a file full of �� where accents should be, or an SRT pasted into Word that still has 900 timecodes in it. Here’s the whole path, including how to convert files you already downloaded.

Two ways to end up with a .txt

• Ask for TXT at extraction time. Caption tools that offer formats usually include plain text; picking it means the timecodes never reach your file. One step, nothing to clean.
• Take the SRT and strip it. Necessary when you already have the subtitle file, when the only track you could get was SRT, or when you want both the timed and untimed versions from one download.

If you haven’t downloaded anything yet, take the first route and skip the conversion section. If you’re here because you grabbed an SRT last week, jump to converting. For the difference between the three output formats and when each belongs, see our guide to timestamped caption extraction; if you want readable prose rather than a file per se, the subtitle text extractor guide covers the clean-up patterns in more detail.

The direct route: download as TXT

1. Open YouTube Caption Downloader and paste the video URL — watch link, youtu.be link or Shorts link.
2. Pick the track. Prefer the manual caption if one is listed: it arrives with punctuation and sane phrasing, so the text file needs almost no editing. Auto-generated and translated tracks work identically; they just need more clean-up afterwards.
3. Choose TXT and download. No account, no install — the site is free with no signup and produces SRT, VTT or TXT from the same track.
4. Rename the file to something you’ll recognise in six months: 2026-10-08-kubecon-keynote.txt beats captions.txt. Keep the video ID or URL in the first line of the file if you might need to re-fetch a better track.

You now have a text file. The rest of this page is about the things that go wrong after that.

Converting an SRT or VTT you already have

An SRT file is four things per cue: an index number, a timecode line with an arrow, the text, and a blank line. Delete the first, second and fourth and you have a transcript. VTT is the same idea with a WEBVTT header, optional cue identifiers and dot-separated milliseconds.

In any editor with regex find-and-replace (VS Code, Notepad++, BBEdit), on a copy of the file:

Find:   ^\d+\s*$            Replace: (empty)     # cue index lines
Find:   ^\d\d:\d\d:\d\d[,.]\d\d\d.*           Replace: (empty)     # timecode lines
Find:   ^WEBVTT.*$          Replace: (empty)     # VTT header/style blocks
Find:   \n\n+               Replace: \n          # collapse blank lines

For a folder of files, a few lines of Python beat clicking:

import pathlib, re
for p in pathlib.Path(".").glob("*.srt"):
    lines = [l for l in p.read_text(encoding="utf-8").splitlines()
             if not re.match(r"^\d+$", l) and "-->" not in l]
    p.with_suffix(".txt").write_text("\n".join(lines).strip() + "\n", encoding="utf-8")

On a Mac with no editor to hand, the shell does it in one pass — grep -v -E '^[0-9]+$|-->|^$' file.srt > file.txt. And if the alternative to converting is re-extracting, the browser tool takes ten seconds per video, which is often the faster answer for one or two files.

Encoding: the accent problem

If your transcript of a Spanish, French, German or Japanese video looks like café instead of café, the download didn’t fail — something opened or re-saved the file with the wrong character encoding. The fix, in order of likelihood:

• Windows Notepad: File → Save as → set Encoding to UTF-8 and overwrite. On older Notepad versions, choose “UTF-8 with BOM” if the plain option still misreads the file — see our note on the BOM for why that exists.
• VS Code: the encoding name sits in the status bar bottom-right. Click it → Reopen with Encoding → UTF-8, then Save with Encoding → UTF-8.
• Excel: importing a .txt is where encoding goes most visibly wrong. Use Import from Text with “Original Data Approach” set to 65001: Unicode (UTF-8) rather than letting it guess.
• Copy-paste through a browser form silently re-encodes. Move files, don’t move clipboard text, when the language has accents.

Cyrillic, Chinese, Arabic and Hebrew survive extraction fine as long as the file stays UTF-8 end to end. Right-to-left scripts can look scrambled in editors that don’t handle bidi text — that’s a display issue, not a corrupt file, and opening it in a proper editor settles it.

Line wraps, paragraph breaks and duplicates

Three cosmetic repairs make an auto-track .txt readable, and they belong in this order:

1. Kill duplicated consecutive lines — the rolling-caption artefact. Regex ^(.+)\n\1$ → $1, run until the count stops falling.
2. Join the fragments into paragraphs: caption cues are one line per breath, so a sentence is often split three ways. Join lines, then break at sentence boundaries or around every 90–110 characters.
3. Separate speakers in interviews. Automatic tracks don’t label who’s talking; a paragraph break per turn is the minimum that makes a conversation readable.

That’s the difference between “a file exists” and “a transcript I can work from”. If your aim is prose rather than a saved file — an article draft, say — our extraction guide covers the repurposing pipeline built on top of this.

.txt, .md or .docx — which to keep

Save the raw capture as .txt and let derived formats be derived. Plain text opens on every machine in perpetuity, adds nothing to the words, and every tool that consumes transcripts later — editors, summarisers, search — takes it. Move to .md when a long transcript needs headings to be navigable, and to .docx when a colleague has to comment on it. Never the reverse: converting .docx back to text is how stray formatting and tracked changes leak into your file.

Two housekeeping habits that pay off: keep the untouched download alongside the edited copy (so you can always tell what the source actually said), and if the text came from someone else’s video, note the URL at the top — caption tracks belong to their creators, and paraphrasing or permission beats republishing. Our tool review and the free video tools category cover the rest of the workflow, including burning a cleaned file back into footage.

Frequently asked questions

How do I save YouTube captions as a plain text file?

Choose TXT when you download the caption track. Paste the video URL into youtube-caption-downloader.com, pick the track you want and select TXT rather than SRT or VTT: the file that lands on your machine contains the spoken words only, with no cue numbers and no timecodes.

My caption text file shows strange characters instead of accents. What happened?

That is an encoding mismatch, not a broken download. Re-open or re-save the file as UTF-8. Windows Notepad defaulting to a legacy codepage, or a copy-paste round trip through a browser form, is what normally turns accents into mojibake. There’s a whole name for it, and a whole fix in the encoding section above.

Can I convert an SRT I already downloaded into a text file?

Yes. Delete every line that is only a cue number, every line containing an arrow timecode, and the blank separator lines. A two-line regular-expression find-and-replace does it, or a short script if you have a folder full of files to process.

Should I save the transcript as .txt, .md or .docx?

Start with .txt because it is the only format every tool on every machine opens without argument. Move to .md when you want headings for a long transcript and .docx when someone else has to edit it — converting later is easy, and going the other way loses formatting.