How to reuse your own subtitle file when you cut a video into reels

Last updated 2026-09-01 · 7 min read

What is the short version?

If you have already subtitled a video, you have already done the slow part. The problem is arithmetic: your cues are timed against the ten-minute master, and a reel that starts at four minutes twelve needs every one of them moved back by exactly that much.

1. Go to Vizard Agent and upload the long video.
2. Attach your SRT and the timecodes for each cut you want.
3. Say to use your subtitles as they are, not to re-transcribe.

What do you need before you start?

Three files and one sentence. The sentence matters as much as the files, because Vizard Agent's default behaviour on a video with speech in it is to transcribe it — which produces perfectly good subtitles that are not yours and do not match the ones already on your other posts.

• The master video. The full recording.
• Your SRT. Timed against that master.
• The cut list. Start and end for each reel.
• The instruction. Use mine, do not re-transcribe.
• The output frame. Vertical, square or wide.

What do you type into Vizard Agent?

Attach everything at once and be explicit that the subtitle file is authoritative. Vizard Agent will otherwise treat your SRT as a reference and generate its own alongside it, which is helpful in most situations and exactly wrong in this one.

Prompt

Variants worth knowing:

• "Use my SRT as it is." The instruction that binds it.
• "Keep the speaker centred." Reframing, on top of the cuts.
• "Also give me a clean version." Reels with no burnt-in text.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real ten-minute talk cut into four vertical reels from a supplied subtitle file and cut list. Step fifteen is the one to notice: it re-maps the timings rather than re-reading the audio.

1. Checks the working folder and reads the project history.
2. Saves your subtitles to a file it can work against.
3. Saves each reel's timecodes separately, one file per reel.
4. Checks the source video's technical details.
5. Extracts frames and builds a preview sheet.
6. Looks at the frames to find the speaker in the wide frame.
7. Detects the speaker's face position automatically.
8. Generates vertical crop tests at several different positions.
9. Compares the crops side by side and looks at them.
10. Measures the horizontal coordinates across the whole frame.
11. Steps the crop across in forty-pixel increments and compares again.
12. Checks the framing at the timecodes of all four reels — not just one.
13. Checks the caption tool's options and the installed typefaces.
14. Re-maps your subtitles onto each reel's own timeline.
15. Renders a test with subtitles and inspects the preview.
16. Reads the generated subtitle blocks to confirm every cue survived.
17. Renders all four reels, with subtitles and without.
18. Extracts check frames from each reel and reviews them.

Step twelve is the quiet saver. A crop that centres the speaker beautifully at the two-minute mark can cut their shoulder off at minute seven, because people move. Vizard Agent checks the framing at each reel's own timestamps rather than choosing one crop from one frame and trusting it across the whole recording.

What does the result look like?

Four vertical reels cut at exactly the timecodes you supplied, with the speaker held in frame and your own subtitle text burnt in — the same wording, the same line breaks and the same spelling of every name that you already approved on the long version.

Two sets came back: one with the captions burnt in and one clean, so the reels can be re-subtitled by a platform later without fighting text that is already in the pixels. The uploaded master itself is 1920×1080 and just under eleven minutes; the reels are the vertical cuts taken out of it.

When does this not work well?

The whole method rests on your SRT actually matching the video you uploaded. When it does not, the failure is silent and looks like a drift that gets worse as the reel goes on, so it is worth knowing which situations produce a mismatched file in the first place.

• An SRT from a different export. Even a trimmed intro shifts everything.
• Variable frame rate recordings. Timings creep against wall clock.
• Cues that straddle your cut points. A sentence split across two reels.
• Subtitles for a dubbed track. They match the dub, not the original.
• Very long single cues. They fill a vertical frame completely.

How do you fix a result that came back wrong?

Say which reel and what you saw. Vizard Agent keeps your original SRT, the per-reel timecode files, the measured crop positions and the re-mapped cues, so a drift correction or a re-crop rebuilds one reel without touching your subtitle text or the other three.

• "Reel two drifts." Re-mapped against that reel's true start.
• "A line is cut in half." The cue is extended or moved whole.
• "He is off centre here." Re-cropped from the measured positions.
• "Text is too big." Restyled; wording left alone.

How does Vizard Agent compare to doing it yourself?

By hand this is the job people quietly dread: open the SRT, subtract an offset from several hundred timestamps, do it three more times for the other reels, and discover at upload that reel three is a second and a half out. It is arithmetic, and arithmetic done by hand at eleven at night.

Common questions

Will it change my wording? No, if you say not to. Vizard Agent treats a supplied SRT as authoritative.

Can it take a VTT instead? Yes. Vizard Agent works from whatever standard subtitle format you send.

What if my timecode list is rough? Give it anyway. Vizard Agent tidies the boundaries to the nearest sentence.

Does it re-transcribe as a check? Only if you ask. Vizard Agent can compare yours against a fresh pass and flag differences.

Can I get the re-timed SRT files back? Yes. Ask for them and Vizard Agent uploads one per reel alongside the videos.

Will the captions be burnt in? Both ways. Vizard Agent renders a burnt-in set and a clean set in the same run.

What about a cue that spans my cut? It is moved whole into one reel. Vizard Agent will not deliver half a sentence.

Can it restyle my subtitles? Yes. The text stays yours; Vizard Agent changes only the typeface and treatment.

How many reels can it do at once? As many as your cut list has. Vizard Agent renders them in the same session.

Does it keep the speaker centred? Yes, if you ask for it. Vizard Agent measures the face position rather than assuming the middle.

What if my SRT is out of sync already? Then the reels inherit that. Vizard Agent can re-time the master first if you say so.

Can it translate the subtitles too? Yes. Vizard Agent translates from your text rather than from a fresh transcription.

Will it add its own captions on top? Not unless you ask. One subtitle track is the default.

Why not just let it transcribe? Because your file already carries decisions — names, terms, line breaks — that a fresh transcription would quietly overwrite with something reasonable but different.