Captions reach a viewer in one of two ways. Either they travel beside the video as a separate text file that the player draws on top, or they are drawn into the picture itself before the video is saved. The first kind comes as SRT or WebVTT. The second is called burned-in, or open, captions. Each solves a problem the other cannot, and the right choice depends almost entirely on where the video is going.
What an SRT file looks like
SRT is the oldest and most widely accepted subtitle format. It is plain text: a number, a time range, the words, and a blank line.
1 00:00:01,200 --> 00:00:03,900 The meeting starts at ten. 2 00:00:04,100 --> 00:00:06,500 Bring the signed forms with you.
Times are hours, minutes, seconds and milliseconds, with a comma before the milliseconds. Many players also understand a few formatting tags such as <i> for italics, but there is no formal standard for them, so do not rely on styling surviving.
What a WebVTT file looks like
WebVTT (.vtt) was designed for web video and looks almost the same:
WEBVTT 00:00:01.200 --> 00:00:03.900 The meeting starts at ten. 00:00:04.100 --> 00:00:06.500 Bring the signed forms with you.
The visible differences are small, and they are exactly the ones that make a file fail to load:
- The file must begin with the line WEBVTT.
- Milliseconds follow a full stop, not a comma.
- Cue numbers are optional.
- The text of a cue may not contain the arrow "-->". The exporters here replace it if your words happen to contain one.
WebVTT can also carry positioning (put this cue at the top, align it left) and can be styled with CSS on a web page. For most people that matters less than the main point: browsers play WebVTT natively and do not play SRT. If you are adding captions to a video on your own website with the HTML track element, you need a .vtt file.
Save both formats as UTF-8, which the tools here always do. That is what lets Hindi, accented letters and other non-English text come through intact instead of as question marks.
Burned-in captions
Burned-in captions are not a file at all. They are pixels in the picture, drawn on every frame before the video is encoded. That has clear advantages:
- They show everywhere the video plays, with no upload step and no setting for the viewer to find.
- They work when a video autoplays with the sound off, which is how many people first see a video in a social feed.
- You control the look completely: font, size, colour, position, and effects such as highlighting the word being said.
And clear costs:
- Nobody can turn them off, resize them or switch to another language.
- A typo is permanent. Fixing it means exporting the video again.
- Platforms and search engines cannot read them as text, so they add nothing to search or automatic translation.
- They can be cropped off when a platform reframes the video, for example from landscape to a vertical feed.
Where each is accepted
| Destination | What to give it |
|---|---|
| YouTube | Upload an SRT or VTT file in YouTube Studio. Viewers can turn them on and off, and the text is indexed. |
| Vimeo, Facebook, LinkedIn | Each accepts an SRT file with the upload. Vimeo also takes WebVTT. |
| Short-video apps and stories | Most give you no place to upload a caption file. Burn the captions in, or use the app's own caption feature. |
| Your own website | WebVTT, through the HTML track element. |
| VLC and most desktop players, many TVs | An SRT file with the same name as the video, in the same folder, is usually picked up automatically. |
| Video editing software | SRT is the format almost every editor imports. |
| WhatsApp, email, presentations | Burn them in. There is nowhere for a separate file to go. |
Platforms change their upload options from time to time, so check the platform's own help page if a file is refused.
For anything that will be shared widely, the usual answer is both: burn the captions into the version you post on social apps, and upload an SRT to YouTube or wherever a caption file is accepted, so viewers who rely on captions can resize them and search engines can read them.
How long a caption should be
Whatever the format, captions fail when there is too much to read in the time given. Professional guidelines are a useful reference. Netflix's English style guide allows at most 42 characters per line, two lines per caption, and a reading speed of up to 20 characters per second for adult programmes. The BBC's guidelines put comfortable reading at 160 to 180 words per minute. Both also keep a caption on screen long enough to be read: well under a second is too short for anyone.
Where a line breaks matters as much as its length. "Bring the signed / forms with you" is harder to read than "Bring the signed forms / with you". A good break falls at a pause or the end of a phrase, never between a word and the small word that belongs to it.
How the tools here break lines and export
Audio to Text saves TXT, SRT and VTT. Whisper hands back pieces of speech with a start and end time. Any piece longer than 84 characters, about two short lines on screen, is split into several captions, and the cut points are chosen together rather than by filling each caption to the limit: after a comma or the end of a clause where possible, at similar lengths, with no one-word stragglers, and never before a word that leans on the one before it, such as है, की or में in Hindi. Each new caption gets a share of the time in proportion to its length.
The video editor works word by word. Its captions start a new line at every pause longer than 0.7 seconds and after the end of a sentence, and keep each line to at most 5 seconds and 32 characters, which suits a phone screen. The same phrase-aware rule chooses where to break between those points, so lines in a song or a fast talk do not end mid-phrase. On the video itself, the look you choose decides how many letters show at once (the Subtitle bar look shows up to 42, the One word look a single word), and the "Most letters at once" slider sets it anywhere from 8 to 60; a longer line is shown in pieces.
Because each caption is tied to the moment in the original clip where its words were spoken, captions follow your edits: cut a pause, move a section or speed the video up, and the words stay on the right frames. The .srt and .vtt downloads are timed to the finished video, after every cut and speed change, and each caption stays up 0.2 seconds after its last word unless the next one starts sooner.
Burned-in captions come from the same editor. You pick a look, then adjust font, size, position (top, middle, bottom, or drag them anywhere on the video), outline, shadow and an optional box behind the text. You can show the whole caption, reveal it word by word, or show one word at a time, and highlight the word being said by colour, karaoke-style fill, a box or a size change. The export draws exactly what the preview shows into the video, so any language and font that displays in your browser comes out the same in the file; Hindi words are drawn in a matching Hindi font when the chosen font has none.
If the video already has words burned in along the bottom, such as lyrics or someone else's subtitles, the editor notices and puts your captions at the top so the two do not overlap.
A short decision guide
- Posting to a social feed or sending on WhatsApp: burn them in.
- Uploading to YouTube, Vimeo or a course platform: upload an SRT, and burn in only if you want a particular look.
- Your own web page: WebVTT.
- Handing the video to an editor or translator: SRT, so they can change the text.
- Not sure: export the SRT anyway. It takes seconds, and you will want it later.
The captions are only as good as the words in them. If yours came from speech recognition, check them first; choosing a speech-to-text model and auto captions for songs cover how to get them right.