Export SRT captions and bring them into Final Cut Pro
Saving a finished transcript as subtitles or plain text, how the blocks are split, and how the timing lines up on import.
A finished transcription offers two exports. Export SRT writes timed subtitle blocks built from word timestamps, saved as your source clip's name with an srt extension. Export Text writes the continuous transcript with a txt extension. A new subtitle block starts whenever the gap from the previous word passes six tenths of a second, the block would run longer than seven seconds, or it already holds twenty words. In Final Cut Pro, choose File, then Import, then Captions. Caption timing matches the file you sent to RoughCut, so an audio-only export from Final Cut Pro must start at the same point on the timeline as the project you are captioning. Ranges the model could not hear cleanly are flagged on screen before you export, and are worth listening to first. SRT is widely accepted elsewhere.
On this page
Prerequisites
- A finished transcription showing the
Your transcript is readybanner. See Transcribe without cutting. - If you are captioning a finished edit, the audio you transcribed must have come from that edit — ideally through an audio-only Final Cut Pro share destination.
Final Cut Pro's menu wording belongs to Apple and changes between versions. Apple's own Final Cut Pro user guide is authoritative. If a label below does not match your copy, follow Apple's wording and look for the equivalent caption import command.
Steps
- Open the finished transcription job. The result screen shows the word count, the duration, and the detected language.
- Read the flagged ranges, if there are any. A heading like
<n> sections flagged — listen before trustinglists each range that could not be transcribed cleanly. - Listen to those ranges in your source media and correct them later if the transcript is wrong there.
- Click
Export SRT…for timed subtitles, orExport Text…for the plain transcript. - Save the file. The default names are your source clip's name with
.srtor.txt— for example,Interview_take3.srt. - In Final Cut Pro, open the project you want to caption and choose
File → Import → Captions. - Select the exported
.srtfile and confirm the import. - Check the first and last caption against the picture. If everything is offset by the same amount, the audio you transcribed did not start where the project starts — see below.
Expected result
A caption track on your Final Cut Pro timeline, one subtitle per block, timed to the audio you transcribed. From there Final Cut Pro's own caption tools take over: editing text, adjusting timing, changing roles, and burning in or exporting a sidecar file.
SRT or plain text?
Export SRT… | Export Text… | |
|---|---|---|
| Contains | Numbered, timed subtitle blocks | The continuous transcript as running text |
| Timing | Yes, from word timestamps | None |
| File | <stem>.srt | <stem>.txt |
| Use it for | Captions, subtitles, anything on a timeline | Show notes, blog drafts, search, quoting yourself |
Both come from the same transcription. Exporting one does not stop you exporting the other.
How the transcript is split into blocks
RoughCut builds subtitle blocks from word-level timestamps rather than from sentences. A new block starts as soon as any one of these is true:
| Rule | Threshold |
|---|---|
| The gap from the previous word is longer than | 0.6 seconds |
| The block would run longer than | 7.0 seconds |
| The block already holds | 20 words |
The practical effect is that blocks break at natural pauses first, and the length limits only take over during continuous speech. Timecodes are written in the standard HH:MM:SS,mmm form.
These are fixed. If you need different pacing — longer blocks for a slow read, shorter ones for a fast one — re-time the captions in your editor after import, where you can see them against the picture.
Where the timing comes from
Caption times are measured from the start of the file you sent to RoughCut, and from nothing else.
That is what makes an audio-only share destination the right way to caption a finished edit: Final Cut Pro exports the timeline's audio from the timeline's start, so the transcript's zero and the project's zero are the same point. The captions land where the words are.
It also explains the two ways this goes wrong:
- Everything is offset by a constant amount. The audio you transcribed began at a different point than the project does — a trimmed export, a range selection, or a different version of the edit. Re-export the audio from the top of the timeline and transcribe that.
- The captions drift progressively. The audio and the project are not the same recording, or not at the same speed. Re-export and transcribe again.
Captioning an edit you have since changed will also misalign it. Transcribe after the cut is locked, not before.
Flagged ranges — listen before you publish
The result screen flags every range the model could not hear cleanly, with a line per range and a note reading either couldn't transcribe cleanly or transcription may have degraded. Nothing was silently dropped; RoughCut tells you where it is unsure instead of quietly guessing.
These ranges export like any other. A caption you never checked is a caption you are publishing on trust, so listen to each flagged range before the file goes anywhere public — that is a minute or two of work against a mistake that stays on screen.
Other apps
SRT is a plain, widely accepted subtitle format. Adobe Premiere Pro, DaVinci Resolve, and YouTube all read it, as do most other editors and platforms.
What each one does with the file is its own business: how blocks are styled, whether they arrive as a caption track or a text layer, and which SRT extensions are honoured all vary by application and version. Expect the words and their times to survive; check the presentation in whichever tool you are using.
Common mistakes
- Transcribing the video render when the audio would do. An audio-only share destination skips the video render and gives identical timing. See Send media from Final Cut Pro to RoughCut.
- Captioning an edit that has changed since you transcribed it. Every change after the transcription shifts everything after it.
- Exporting
.txtand expecting captions. Only the SRT carries timing. - Publishing without checking the flagged ranges. They are flagged precisely because the words may be wrong.
- Editing an SRT by hand to fix timing. Import it and adjust in the editor, where you can see the result.
Troubleshooting
SRT file is missing from the job folder. The transcript file for that job is not on disk — most often because the job folder was cleaned up. Transcribe the file again.
Could not load the transcript: … The job's transcript could not be read. Re-run the transcription; if it fails again, see Troubleshooting by symptom.
Final Cut Pro imports the SRT but shows nothing. Confirm you used the caption import command rather than a general file import, and check that the caption track is visible in the timeline.
The text is right but the language is wrong. The detected language is shown on the result screen. If it guessed wrong, transcribe again with Language set explicitly instead of Auto-detect. See Switch between English and Arabic.