Why your downloaded video is blurry even at 1080p: bitrate explained
A 1080p file can still look soft. We scored eight copies of one clip: bitrate decided sharpness as much as size. How to check yours and what helps.
SRT vs VTT vs TXT: what is inside each subtitle file, where each one works, how to convert them, and why YouTube's automatic captions repeat every line.

Choose SRT for video editors and uploads, VTT for captions on your own web page, and TXT when you only want to read the words. SRT and VTT (WebVTT) both hold timed captions; VTT adds a header, styling and positioning and is the only format a web page's video player reads. TXT has no timings.
You have a video and you need its words: as subtitles for an edit, as captions on a web page, or as text to read. Three files come up again and again: SRT, VTT and TXT. This page compares SRT vs VTT vs TXT in plain words. It shows what is inside each file, where each one works, how to change one into another, and what happens when the captions come from YouTube's automatic speech recognition. That last part is the one most guides leave out. We measured it on five real talks on 26 October 2026.
Pick the file by what you will do with it. The table gives the short answer. The sections below explain why.
| You want to | Take | Why |
|---|---|---|
| Put subtitles on a video in an editor (CapCut, Premiere Pro) | SRT | Editors import it. CapCut's help page recommends it |
| Upload captions to YouTube | SRT or VTT | YouTube lists both. VTT keeps position and bold, italic and underline |
| Show captions in a video on your own website | VTT | A web page's <track> element reads WebVTT files |
| Read, search, quote or translate what was said | TXT | Plain text with no numbers or times in the way |
| Know when something was said | SRT | Every caption has a start and an end time |
| Let YouTube set the timings for you | TXT | YouTube can match a plain transcript to the speech |
If you are not sure, take SRT. It is the simplest timed format, and turning it into VTT later takes a minute and loses nothing.
All three are plain text files that you can open in Notepad or TextEdit. The difference is how much they say about time and looks. Here is the same caption in each one.
1
00:00:27,103 --> 00:00:29,678
Good morning. How are you?
An SRT file is a list of numbered captions. Each one has a number, a line with the start and end time, the text, and an empty line. The times use hours, minutes, seconds and a comma before the milliseconds. The Library of Congress gives the format as hours:minutes:seconds,milliseconds (00:00:00,000). SRT began as the output of SubRip, a program that pulled subtitles off DVDs. It never got a formal standard: the Library of Congress notes "little to no information on SRT formal documentation", and it says SRT files "do not have a character encoding standard".
WEBVTT
00:00:27.103 --> 00:00:29.678
Good morning. How are you?
VTT is short for WebVTT, the Web Video Text Tracks format. It is a W3C standard, last published as a Candidate Recommendation Draft on 20 May 2026. The file must start with WEBVTT. The times use a dot before the milliseconds, and the hours can be left out when they are zero. The numbers above captions are optional. The standard also says the file "must" be encoded as UTF-8. VTT can do what SRT cannot: place a caption on the screen, style it, add NOTE comments that players ignore, and hold chapter titles.
Good morning. How are you?
A TXT transcript is only the words. It has no numbers and no times, so no player can show it as subtitles. In exchange it is the easiest to read, search and paste into a translator, a summary tool or your notes. It is also the smallest file. From a 20-minute TED talk, our TXT was 18 KB, our SRT 32 KB.
SRT and VTT carry the same thing, timed lines of text, so most files can move from one to the other. The differences are in the details. The VTT column comes from the W3C standard. The SRT column comes from the Library of Congress description and YouTube's help page.
| SRT | VTT (WebVTT) | |
|---|---|---|
| First line | The number 1 | WEBVTT (required) |
| Time format | 00:01:02,500 (comma) | 00:01:02.500 or 01:02.500 (dot; hours optional) |
| Numbers above captions | Expected | Optional, as cue names |
| Position on screen | No standard way | Yes: line, position, size, align, vertical, region |
| Styling | Bold, italic, underline and colour tags, shown by some players only; YouTube reads none | STYLE blocks and inline tags; YouTube reads bold, italic, underline |
| Comments | No | NOTE blocks, ignored by players |
| Encoding | No rule; old files may not be UTF-8 | UTF-8 only |
Web pages (<track>) | Not read | The format <track> uses |
| File type label | text/plain | text/vtt |
If you only need words and times, both do the job. VTT matters when captions must sit somewhere else on the screen, for example above a name shown at the bottom, or when they go into a web page.
For a classroom with no internet, see how to show a video and its captions offline.
The best format is the one your next program accepts. These are the rules from each program's own help pages, read on 26 October 2026 (LinkedIn's on 28 October).
| Where the file goes | SRT | VTT | TXT (no timings) |
|---|---|---|---|
| YouTube upload | Yes. Basic SRT, "No style info (markup) is recognized", UTF-8 | Yes. "In initial implementation": position, plus bold, italic and underline | Yes, as Without timing: YouTube sets the times |
| A video on your own web page | No | Yes | No |
| CapCut on a computer | Yes, recommended | Not listed | Only "TXT with timecodes" |
| CapCut on the web | Yes | Not listed | "TXT files are not supported on Web" |
| CapCut phone app | No subtitle file import "as of 2026" | No | No |
| Premiere Pro | Yes, import the SRT | Not checked | No |
| LinkedIn video post | Yes, the only caption file its help names, attached on a computer | Not listed | Not listed |
A few points the table cannot hold:
.sbv), TTML, SCC and others. SBV is YouTube's older format with a different time style. You rarely need it.Most subtitle files people download come from YouTube, and most YouTube captions are made automatically from speech. Those files are not like the clean samples above. This is the part the other comparisons miss, and it decides whether your SRT works.
YouTube shows automatic captions as two rolling lines: a new line appears at the bottom while the old one moves up. The VTT file is written to play that animation. Here is the start of one, from Julian Treasure's TED talk:
00:00:14.080 --> 00:00:16.870 align:start position:0%
the<00:00:14.240><c> human</c><00:00:14.920><c> voice</c><00:00:15.920><c> it's</c>...
00:00:16.870 --> 00:00:16.880 align:start position:0%
the human voice it's the instrument we
Every word carries its own time tag. Then a caption that lasts 10 milliseconds repeats the whole line, so the line can scroll up. The next caption shows it again above the new words. We counted this in five public talks, with no sign-in, on 26 October 2026:
| Talk (length) | Captions in the automatic VTT | Lines that repeat the line before | Words in a plain copy of the VTT vs words spoken | YouTube's own SRT: captions that overlap the one before |
|---|---|---|---|---|
| Sir Ken Robinson, TED (20:03) | 1,139 | 1,110 | 9,819 vs 3,337 | 555 of 584 |
| Tim Urban, TED (14:04) | 657 | 644 | 6,648 vs 2,256 | 322 of 335 |
| Steve Jobs, Stanford (15:04) | 672 | 662 | 6,872 vs 2,314 | 331 of 341 |
| Amy Cuddy, TED (21:03) | 1,064 | 1,058 | 11,458 vs 3,832 | 529 of 535 |
| Julian Treasure, TED (9:58) | 559 | 550 | 4,868 vs 1,646 | 275 of 284 |
Three things follow from these numbers:
A clean SRT from automatic captions needs two fixes: drop each line that only repeats the one before, and end each caption where the next one starts. Here is Tim Urban's opening in YouTube's SRT. The first caption is still on screen until 17 seconds, long after the second one appears:
1
00:00:12,559 --> 00:00:17,039
so in
2
00:00:14,360 --> 00:00:19,640
college I was a government major which
And the same opening in the SRT our downloader makes from the VTT. Each caption ends before the next begins:
1
00:00:12,559 --> 00:00:14,350
so in
2
00:00:14,360 --> 00:00:17,029
college I was a government major which
Same words, same number of captions, but in all five talks ours had no caption overlapping another. Automatic captions also have no full stops. Our measurement found 0 to 3 in a whole talk. So the TXT has to be broken into paragraphs some other way, or it arrives as one block of 3,000 words. Our TXT now starts a new paragraph every eight lines when there is no full stop to break on.
You can change one file into another with any text editor. Do not only rename the file: a VTT without its first line is not a VTT file, and the times are written differently.
| From → to | What to change | What you lose |
|---|---|---|
| SRT → VTT | Add WEBVTT and an empty line at the top. Change the comma to a dot in every timing line | Nothing |
| VTT → SRT | Delete the WEBVTT line, any NOTE, STYLE and REGION blocks, and the settings after the times (align:start position:0%). Remove tags such as <c>. Number the captions. Change the dot to a comma, and add 00: where the hours are missing | Position and styling |
| SRT or VTT → TXT | Delete the numbers, the timing lines and the empty lines | All timing |
| TXT → SRT or VTT | Not possible by hand without the times. Upload it to YouTube Without timing, let YouTube set the times, then download the result | Nothing, if YouTube's matching is right |
Change commas only in the timing lines, not in the text. A search-and-replace over the whole file would also change "Well, hello" into "Well. hello". If the file came from YouTube's automatic captions, convert it only after the repeats are removed, or you carry them into the new file.
Raptor Downloader saves a video's captions as TXT or SRT. For YouTube's automatic captions, it removes the rolling repeats and keeps each caption's own end time, so captions do not overlap. It works for public videos that have captions, on YouTube and on other sites that offer them, such as Dailymotion and Facebook.
en.We list up to three caption languages for each video: the creator's captions first, then the automatic track in the language spoken. If a video has no captions in any language, you see "That video has no captions in any language, so there are no words to save." Its video and sound still download. Files are saved as UTF-8. Each download button works for 30 minutes after the link is read.
The YouTube transcript downloader page does the same for one YouTube video and explains its caption options. For a whole channel or playlist, the YouTube channel downloader saves the TXT or SRT of every video you tick. The guide to sound, words and subtitles shows the buttons step by step. We do not offer a VTT button: if you need VTT for a web page, convert our SRT with the two changes in the table above.
A subtitle file only exists when someone made captions as separate text. Some captions are part of the picture instead, burned into the video. No tool can save those as SRT, because there is no text to save, only pixels. Twitch clips are a common case, as our article on Twitch clips and their captions explains.
We only read captions that anyone can see without signing in. We never get around logins, private settings or paywalls. Bilibili, for example, shows subtitles only to signed-in users, so we save none from it.
The words in a transcript belong to the person who said or wrote them. Use transcripts for notes, study, captions on your own videos and short quotes with credit. Publishing someone else's full transcript needs their permission. Our guide on whether downloading a video is legal explains the rules. This is general information, not legal advice.
Only for the web. VTT can place and style captions and is the format a web page's video player reads. For an editor, an upload or a media player, SRT is the safer choice, because more programs import it. YouTube accepts both.
A VTT file starts with the word WEBVTT, uses a dot before the milliseconds (00:00:01.500) and may carry position, styling and comments. An SRT file has numbered captions, a comma before the milliseconds (00:00:01,500) and no standard styling. The words and the timings are the same.
No. A browser rejects a VTT file that does not start with WEBVTT, and the commas in the times must become dots. Add the WEBVTT line and change the commas in the timing lines, and the file works. Going the other way, you also remove the VTT-only lines.
Both. YouTube's help page lists SubRip (.srt), with no styling recognised, and WebVTT (.vtt), with positioning plus bold, italic and underline. Both must be saved as UTF-8. YouTube also takes a plain transcript without timings and sets the times itself.
Yes, and nothing is lost. Put WEBVTT and an empty line at the top, then change the comma before the milliseconds to a dot in every timing line. The numbers above each caption can stay. VTT to SRT loses positioning and styling, which SRT cannot hold.
UTF-8. The WebVTT standard requires it, YouTube requires it for SRT, and CapCut recommends it. SRT itself has no encoding rule, so an old SRT may be in another encoding and show broken accents. Open it in a text editor and save it again as UTF-8.
The captions came from YouTube's automatic track, which rolls two lines at a time. YouTube's SRT of that track has no repeated text, but almost every caption overlaps the next, so two show at once. Use an SRT whose captions end before the next one starts.
Yes. In YouTube Studio choose Upload file, then Without timing, and YouTube matches the words to the speech. YouTube says the transcript must be in the language spoken in the video, and it does not recommend this for videos over an hour or with poor sound.
A 1080p file can still look soft. We scored eight copies of one clip: bitrate decided sharpness as much as size. How to check yours and what helps.
HLS and DASH cut a video into short pieces and a playlist. What that means when you save one, with the pieces counted on real public links.
MP4 is far smaller and sharper than GIF for the same loop. See our measured sizes, the 256-colour limit, and the few places that still need a real .gif.
A download is never sharper than the biggest copy the site kept. Measured on 56 videos: why you get 720p, what "4K" tools give you, and how to get the best.
First published . Written and checked by the Raptor Downloader team (how we write and check guides). Download only videos you own or have permission to use. Terms · Copyright and takedowns