What the SRT file looks like
An SRT file is plain text. Each cue has a number, a start and end time, and the line to show. This is an example of the layout the page writes:
1
00:00:01,200 --> 00:00:04,000
Welcome back to the workshop.
2
00:00:04,300 --> 00:00:07,100
Today we fit the new door hinge.
3
00:00:07,400 --> 00:00:10,200
Make sure the hinge is lined up.
Times run hours:minutes:seconds,milliseconds, with a comma before the milliseconds. A blank line separates the cues. The page writes each cue as a single line and leaves it to the player to wrap long text.
SRT or VTT
Both hold the same cues. SRT is the older format and the one most video editors, players and upload forms ask for. VTT starts with the word WEBVTT, uses a dot instead of a comma in the times and has no cue numbers. It is the format for subtitles on a web page. Take whichever one the place you're uploading to asks for. On this page the SRT button comes first.
Timing is per segment
Whisper splits speech into segments and times each one. Each segment becomes one cue, so a cue is whatever Whisper treated as one stretch of speech, and it can run longer than you would normally put on screen. The page doesn't time single words, and it doesn't split or merge cues for you. If a cue is too long or starts a beat late, fix it in the file.
Fixing a cue by hand
Edits in the text box change what Copy and TXT give you. They don't reach the SRT or VTT file, which is built from Whisper's segments. To correct a name or a mistake in a cue, download the .srt, open it in any plain text editor, and change the line under that cue. Leave the numbers and the times alone unless you mean to retime it. Save it as UTF-8 text, keeping the .srt ending.
Subtitles in English from another language
Above the file box there is a tick box, "Translate the result to English". With it ticked before you pick the file, Whisper writes the spoken words as English text, and the SRT then holds English lines. It only goes into English. For subtitles in any other language, take the transcript in the original language and translate the lines yourself.
Limits
Audio and video both work. In Chrome and Edge, a video of several GB works on a computer. Browsers that can't read a file in pieces stop at 1 GB on a computer and 200 MB on a phone. The page takes a file, not a web link. Whisper doesn't tell speakers apart, so cues carry no names.
More ways to use it
Welcome back to the workshop.