suubbs is a subtitle editor for Mac that can also write the first draft for you. Open an SRT or VTT you already have, or generate one from the video, then fix words, retime and re-split lines, and style them against a preview that matches the export. It saves SRT, VTT, styled ASS and word-level JSON. The Mac app is a $9.99 one-time purchase for Intel and Apple Silicon on macOS 11 or later, and the same editor runs free in your browser on the first 60 seconds of a video.
Your video on the left, your lines in the middle, your styling on the right — and each of these jobs one click away rather than four menus deep.
An SRT is plain text, so in principle TextEdit can edit one. In practice every change to a timestamp means scrubbing the video, reading off a time and typing it to the millisecond, and every change to a line's length means rebalancing the lines around it by hand. A subtitle editor exists to do four things that a text editor cannot:
Most video editors treat subtitles as a checkbox. suubbs treats them as the entire product, which mostly means the tools you need are visible and the ones you do not need are absent. And it is a native Mac app — if you arrived looking for Subtitle Edit, which Mac users have had to run through Wine or a Windows virtual machine, suubbs vs Subtitle Edit says honestly which of its jobs suubbs covers and which it does not.
You do not have to transcribe anything to use the editor.
.srt or .vtt.
A .json exported from suubbs reopens with everything it had: words, word
timings and styling.One honest limitation. An SRT or VTT only says when each line starts and ends, so on import suubbs spreads the words evenly across each line. Everything line-level works exactly as on a generated transcript. Word-level highlighting on an imported file is approximate, though, because the file never contained the word timings. If you need those, generate from the video instead: the Mac app measures word boundaries to about 3 ms.
If the subtitles you want to edit are inside a video file as a track, extract them first; the result opens here like any other SRT. If all you need is the other format, converting SRT to VTT does not need an editor at all. And if you have no file yet, making an SRT file from a video walks through generating one.
Every word carries its own timestamp, and a line's text is built from its words rather than stored beside them. That one decision is why nothing below breaks anything else.
Every transcript has three bad words in it — a name, a piece of jargon, something swallowed by a passing bus. Click the line, retype the word, and the line keeps its timing; the words around it keep their measured timestamps, and only the corrected word's time is estimated. There is no dialog and no re-sync step. Improving subtitle accuracy explains exactly what an edit does to the timing. In the Mac app you can prevent most of them up front: type the names and jargon into Names and jargon before you press Generate, and the model is biased toward your spellings.
Each line in the list has its start and end time as editable fields: type a time and press Enter. Or drag the line on the timeline, which sits over the audio waveform so you can see where speech starts. The quickest way is the keyboard: play to where the line should begin and press [, play to where it should end and press ]. A line can never be dragged shorter than 100 ms or turned inside out.
Lines sit on layers, so two can overlap without one being shoved out of the way: a speaker label above dialogue, a description of a sound while someone talks, a sign being read out. That is also what makes it practical to write closed captions rather than plain subtitles, which need room for exactly those non-speech lines. How each export carries two lines at once, and the one burn-in case where they print over each other, is on overlapping subtitles.
| Key | What it does |
|---|---|
| Space | Play and pause |
| ← / → | Step one frame back or forward |
| Shift ← / → | Jump one second |
| [ / ] | Set the selected line's start or end to the playhead |
| S | Split the selected line at the playhead |
| M | Merge the selected line with the next one |
| N | Insert a new line at the playhead |
| Delete | Remove the selected line |
| ⌘Z / ⇧⌘Z | Undo and redo |
| ? | Settings and shortcuts |
Some fixes are about the file, not a line. They live in Timing & layout, in the ··· menu.
A file that runs early, late or drifts against the video is its own job, with its own diagnosis — see fixing subtitles that are out of sync.
Because the words are the source of truth, the lines can be rebuilt at any point.
Choose a line layout in Timing & layout — Widescreen, 42 characters, 2 lines or Vertical / social, 22 characters, 3 lines — or set Words on screen low for word-at-a-time social captions. The lines re-split as you change it, and anything you have already edited keeps its own text and timing. Change your mind about line length after an hour of editing and you keep the hour.
A line that reads faster than 20 characters a second, or runs past the line limit, is marked in the list with the reason. Those limits, and why they are the usual ones, are in subtitle reading speed.
The preview and the exported file are built from the same document, so they agree.
Start from a preset — Classic, Boxed, Pop, Highlight or Film — and tune font, size, colour, letter spacing, background, position and word-level highlighting on the footage. Save a look you like as your own preset. Style the whole document, one line or one word. How big to make the type, and what each format does with that size, is on subtitle font size; for vertical video, the safe-area guide shows where TikTok, Reels and Shorts cover the frame.
When a format cannot carry something you used, suubbs names it as you export, rather than leaving you to find out after the upload.
Including the two things people are most often surprised by.
Yes. Bring in an existing SRT or VTT and edit, restyle and re-export it. You do not have to transcribe from scratch to use the editor. Word timings inside each line are spread evenly on import, because the file never had them.
Yes, in TextEdit after Format ▸ Make Plain Text, as long as you keep the numbering, the comma before the milliseconds and the blank line between cues. Fixing a typo that way is fine. Retiming or re-splitting lines by hand is where an editor saves the time.
Not at line level: the line keeps its start and end, and the words you leave alone keep their measured timestamps. The corrected word itself gets an estimated time within the line, which only shows in a word-by-word highlight.
Yes. Re-segmenting reflows lines against reading-speed and line-length rules without discarding the corrections you have already made.
For ASS, yes — the caption style lives in the document rather than in CSS, and it is the single source for both the preview and the export, so what you see is what libass burns in, with two current exceptions: a panel behind the whole caption needs a one-line ffmpeg override to appear, and a position set on a single line is not exported. SRT and VTT carry only part of it, and suubbs says which part is missing when you export.
Not yet. Export a styled .ass file and ffmpeg burns it into the picture in one command — see burning subtitles into a video. A one-click burn-in is not in the app today.
Intel and Apple Silicon, on macOS 11 or later. It is a native app, with no Wine layer or virtual machine, and once the speech model has downloaded on first launch it works offline.
The Mac app is comfortable with long files. The cue list is not virtualised yet, so a multi-hour transcript with thousands of cues will feel heavy to scroll.
[click the word, type the word]