Export the caption file, not the video
CapCut’s auto captions are drawn to fit a vertical or horizontal frame. Lines are short, punctuation is
light, and the thing you can post is an MP4. A document needs the text track beside that video. On desktop
(Windows or macOS) and on CapCut web, that track is an export option in the same area as video export, not
a hidden developer menu.
Open the project, then Export. In the dialog, find the captions section. CapCut’s own help
tells you to select caption export rather than video export when the file you want is subtitles, and to
choose SRT or TXT. Naming in the dialog moves between versions — “Captions”,
“Caption export”, a checkbox next to the video settings — so trust the section that offers SRT, not a
screenshot from last year. If you want only the subtitle file, turn video export off when the dialog lets
you. You do not need to re-encode the film to get the words.
Official notes are strict about which captions qualify: text that came from auto caption recognition.
Manually typed title cards and some imported layers may not ride along in that SRT. If the export is empty
or missing the control, check that the words on screen were generated as captions and sit on the caption
track. Update the app if the control is absent on a build CapCut’s help says should have it. A Pro mark
beside the control is a CapCut packaging choice. This converter does not sell that export and cannot switch
it on. Generate or obtain the SRT by whatever account you already use, then convert the file locally.
The phone apps are a weak place to finish this job. CapCut’s help describes caption export on mobile in
some versions and, separately, says mobile does not import external SRT the way desktop does. If your phone
build has no SRT or TXT caption export, open the same project on desktop or web and export there. Do not
transcribe the exported MP4 again just because the phone saved a video.
SRT versus the TXT option
Pick SRT when it is offered. SRT has cue numbers and --> clocks, which is
what the converter reads into paragraphs with optional time ranges. TXT from CapCut is sometimes a timed
subtitle dump and sometimes plain lines. If the preview on the converter says no cues were found, the TXT
had no cue structure. Go back and export SRT. Do not spend time adding arrows by hand unless the SRT option
failed and the TXT is all you will get.
CapCut’s help also mentions encoding and export paths: if characters break, the file may need to be UTF-8,
and exotic folders can interfere with their exporter. The practical check is local. Open the SRT in a text
editor. If names and apostrophes look right, drop it on the
SRT to Word converter. If they look like à sequences, re-save as UTF-8 before
converting. The parser removes a UTF-8 BOM when one is present. It cannot repair a file that was saved in
the wrong code page.
What those short lines become in Word
A CapCut caption line is often three or four words because that is what fits over a face on a phone. Spoken
language is not. With Merge same speaker on, consecutive unlabeled cues join when the gap
is about 2.5 seconds or less. A run of chopped lines becomes a paragraph. That is the useful default for a
talking-head video with one speaker.
Two problems survive the merge. Auto captions still mis-hear names and drop punctuation, so the paragraph
is editable, not finished. And a jump cut can leave a gap longer than a couple of seconds, which starts a
new paragraph even though the same person is still talking. That break is usually fine. If the preview
shows a new paragraph every line, the gaps in the SRT are large or merge is off. Turn merge on and look
again before you download.
Leave Show timestamps on if you will quote the video and need to find the moment. Turn
them off for a caption list you are turning into a blog draft or a client script. Download a second file
rather than deleting ranges later. CapCut’s clocks are the editor’s clocks. If they drift against the
exported video, fix them in CapCut or in a subtitle editor and export the SRT again.
Fix SRT timing describes what a Word file will not resync.
Music and trend-audio videos often caption the lyric or caption nothing useful. Delete those cues in the
SRT before you merge, or they become sentences in the document. The same cleanup shows up in other
automatic SRTs; the Whisper-oriented checklist is
Whisper SRT to Word if your file did not come from CapCut’s recognizer
but has the same repeated tails.
YouTube is a different download
Creators often burn CapCut captions into the video, upload to YouTube, and then try to recover words from
YouTube’s auto-captions. You now have two transcripts that will not match. If you still have the CapCut
project, export its SRT and ignore YouTube’s track. If the project is gone and the video is on your
channel, YouTube Studio can download the caption track YouTube stored.
YouTube captions to Word is that path, including why
.sbv is the wrong format to pick. Do not run both files through the converter and try to
reconcile them line by line unless a compliance review forces you to.
Keep the SRT if the video still needs captions
The .docx is for editing and sharing prose. CapCut will not import it back onto the timeline.
Desktop and web can import an SRT or a timed TXT (Captions → Add captions → Import file in
current desktop builds). Mobile, as of CapCut’s 2026 help, does not import that file directly; the
workaround they document is to import on desktop or web and sync the project. So: edit wording that must
remain captions inside the SRT, edit wording that is a blog or a script inside Word, and do not expect one
save to update both.
A single speaker and a clean SRT is enough for this page. If the video is an interview and nobody was
labeled, the paragraphs will read as one voice until you add names.
Interview SRT to Word is the labeling pass. Most CapCut talking-head
exports skip it because there is only one person on camera.
Related workflows
Download the .docx
For a one-speaker CapCut video, merge on joins lines within about 2.5 seconds. Drop the SRT after you have seen a --> line. Download capcut.docx and keep the SRT if the words still belong on the video.