How to make SRT subtitles for a video, free

What an SRT file is

SRT (SubRip Text) is the most widely accepted subtitle format. It is a plain text file: you could open one in any text editor. Each subtitle, called a cue, has three parts and a blank line after it:

1. a number, counting up from 1;
2. a start and end time in hours, minutes, seconds and milliseconds, with an arrow between them;
3. one or two lines of text.

Here is the start of a real file that Transcribe saved from a short podcast clip:

1
00:00:00,000 --> 00:00:02,029
Welcome back to Harbour
Lights, the podcast

2
00:00:02,029 --> 00:00:04,200
about the people who
keep Sunset Bay running.

The same cues as WebVTT, the format web players use, differ in only two ways: the file begins with the word WEBVTT, and the milliseconds come after a full stop instead of a comma (00:00:02.029). The cue numbers are optional in VTT, so Transcribe leaves them out.

Make one, step by step

Typing subtitles by hand takes several minutes for every minute of video. Letting a speech model write the first draft and correcting it is far quicker.

1. Open the video in Transcribe. MP4, MOV and WebM work, and so do audio files (MP3, M4A, WAV and more) if you already have the soundtrack on its own.
2. Pick the language, or leave it on Auto. Setting it helps most on short clips, where detection has little to go on.
3. Transcribe. The first time, the Fast model downloads once (about 77 MB) and is kept by your browser. Long videos are cut at quiet moments into parts of about 30 seconds, so lines appear as they are done.
4. Correct it while it plays. Tap a line's time to hear that moment, and type to fix a word. Use Find and replace for a name the model spelled wrongly, and the line menu to split a line at the cursor or join it with the next.
5. Save the subtitles. Choose Export, then SRT subtitles or WebVTT subtitles. The file downloads straight to your device.

Everything is free with the Fast model, whatever the length. If the speakers have strong accents, there is background noise, or the subject is full of specialist words, Transcribe Pro adds the Accurate model (Whisper small, about 250 MB), which is slower but makes fewer mistakes. For a step-by-step on getting a plain transcript instead of subtitles, see how to transcribe audio to text.

Subtitles people can read

A transcript and a subtitle file hold the same words, but subtitles have to be read in the time they are on screen. The usual rules, which professional subtitlers broadly agree on:

• Two lines at most, each up to about 42 characters, so the eye can take them in at a glance.
• A few seconds per cue. Long enough to read twice, short enough to stay with the speaker; about six seconds is a sensible ceiling.
• Break at natural pauses: after a comma or between phrases, not between "the" and its noun.
• Keep the two lines a similar length where you can. A short line over a long one is harder to read.

Transcribe applies these when it saves: each subtitle is cut to at most two lines of about 42 characters and no more than six seconds, split at the best break near the middle. You can see it in the example above, where one long sentence became two cues. If a cue still breaks somewhere awkward, split or join the lines in the editor and export again.

Adding the file to your video

Name the file after the video, ideally with the language (for example interview.en.srt), so it is easy to keep track of when you make more than one.

• YouTube. In YouTube Studio, open Subtitles, choose the video and its language, then upload the file and pick the option that says it includes timing.
• Other video sites. Look for Subtitles or Captions in the video's settings and upload the .srt there.
• Video editors. Import the .srt like any other file and drop it on the timeline as a captions or subtitle track. Most editors then let you style it.
• Your own website. Add the .vtt with a <track kind="subtitles" srclang="en" src="interview.en.vtt"> inside the <video> element.
• Watching locally. Many desktop players load subtitles automatically when the .srt sits in the same folder with the same name as the video.

SRT file or burned-in captions?

An SRT file gives closed captions: viewers turn them on and off, the platform can search them, and you can fix a typo without exporting the video again. Burned-in (open) captions are part of the picture, so they show everywhere, including on apps and players that ignore subtitle files.

Transcribe Pro can make the burned-in version too: the video plays through once while a captioned copy is recorded on your device, in a Clean, Minimal or Bold word-by-word look, at the bottom or middle of the frame, as filmed or cropped to 9:16 or 1:1. A one-minute video takes about a minute.

Common mistakes

• Uploading the first draft. Speech models mishear names, numbers and quiet words. Watch the video through once with the subtitles before publishing.
• The wrong language set. A recording transcribed as the wrong language comes out as nonsense or as a rough translation. Set it by hand if Auto guesses wrong.
• Saving from a word processor. Rich-text editors add formatting and can change the encoding. Edit subtitles in a subtitle tool or a plain text editor, and keep the file as UTF-8.
• Subtitles for a different cut. If you trim the start of the video after making the file, every cue will be early. Make the subtitles from the final edit.
• Expecting speaker names. Transcribe does not tell voices apart. If it matters who is talking, type the name at the start of the line.