An order comes back as a subtitle file, as both file formats, or as a copy of your video with the captions in the picture. What follows is what is inside each of them, and the conventions the text itself follows.
.srt
The plainest subtitle format there is: a numbered cue, a start and end time, the lines, a blank line.
.vtt
The web format. It opens with a WEBVTT line, separates seconds from milliseconds with a full stop, and can carry cue settings.
.mp4
A copy of your video re-rendered with the captions drawn into the picture, in the style agreed for the order.
Asking for both file formats in one order is normal and barely moves an estimate: they come out of the same timing pass. A burned-in video is a render on top of that pass, which is why it is quoted separately from the files.
That is the whole format. Everything else about a subtitle file is craft rather than syntax: where the text is cut, how long each cue holds, and how the spellings were arrived at.
1 00:00:02,000 --> 00:00:06,240 Every caption starts with the same question: can it be read in time? 2 00:00:07,200 --> 00:00:10,960 Two lines at most, and the second line shorter than the first. 3 00:00:11,200 --> 00:00:14,960 A cue holds long enough to be read, then gets out of the way. 4 00:00:15,320 --> 00:00:18,560 - And when two people speak, - each one takes a line. 5 00:00:18,880 --> 00:00:22,920 Names and technical terms are spelled the way you send them to us. 6 00:00:23,200 --> 00:00:27,200 Sound that carries meaning but no words is described in brackets. 7 00:00:27,440 --> 00:00:28,960 [rain against the window] 8 00:00:29,320 --> 00:00:34,120 Then every cue is checked against the recording before the file is handed over.
A third line takes the eye off the picture. Where a sentence will not fit in two, it is split across cues at a point where the sense holds.
Cue length follows how much text is in it, at a reading speed that suits the audience. A cue that lingers after the words have stopped reads as a mistake.
Cues do not butt up against each other. A short gap is what tells the eye that something has changed, and it is why captions stop flickering.
Marked with a dash, or with names where your platform shows them. Whichever is used is used consistently across the whole file.
Where a sound matters to what is happening and has no words, it is described in square brackets. Where it does not matter, it is left out rather than cluttering the screen.
Names, places, products and terms are set the way you send them, and the same way every time they appear.
A WebVTT file of the same track carries the same cues, opens with a WEBVTT line and writes a full stop where SRT writes a comma. On the home page the two listings can be switched between, over the same timings.
The plain style is white text in a dark box, bottom centred, two lines at most. It is quiet on purpose: on a subtitle file most of it is decided by the player anyway, and on a burned-in video the picture is what people came for.
Where your organisation has a style guide, send it with the video and the burned-in render follows it. Setting one up is work on the first order and nothing on the ones after it, because the settings are kept for you.
On an SRT file most of this is a request to the player rather than a setting in the file. WebVTT carries more of it, and a burned-in video carries all of it, because there the styling is in the picture.
If one of those is what you need, say so anyway: we will tell you plainly whether it is something we can take on, rather than accepting an order and finding out afterwards. Write to operations@subtitleflow.org.