Reuse an SRT file for a shorter video clip

Keep the excerpt, then reset its clock

To reuse a full recording’s SRT for a shorter clip, first identify the exact interval you kept. Remove cues wholly outside that interval from a working copy, review every cue crossing either cut, and subtract the clip’s original start time from the surviving timestamps. Finally, play the exported subtitles with the actual short video. Moving the whole original file earlier is not enough.

This workflow is for one continuous excerpt played at its original speed, with its first retained frame at the new video’s zero point. Use recordings and captions you are authorized to edit and distribute. A montage, reordered scenes, a speed change or an added opening slate needs a different time mapping. Finish the video edit before preparing its delivery subtitles.

Write down exactly what the video contains

Keep the full video and its reviewed subtitles untouched. Record the full-video revision, the short-video revision, and the excerpt’s in and out points measured from the full video’s beginning. Use elapsed hours, minutes, seconds and milliseconds. If your editor displays frame timecode, convert using the actual project timebase; do not read frame digits as milliseconds.

For the original example below, the short video contains the interval from 00:00:12,000 up to 00:00:22,000 of a fictional craft demonstration. It is ten seconds long. There are no internal cuts, speed changes or added titles. The times and words are invented teaching material, not a recording or a Polyglot processing test.

Check the exported video itself. Its audible opening and closing should match the interval you recorded. If the export begins at a different point, correct the reference before changing subtitles. An accurate calculation against the wrong video is still wrong.

Classify cues before deleting or shifting

In a copy of the SRT, mark each cue as outside, fully inside, or crossing a boundary. For this example, a cue ending at or before 12 seconds is outside; one starting at or after 22 seconds is outside. A cue starting before 22 and ending after 12 has some time in the excerpt. This interval test finds candidates to review; caption display time is not proof of which words remain audible.

Keep a small mapping from original cue IDs to the clip’s final cue IDs. Delete only confirmed outside cues from the copy. Retain fully inside cues for playback review, and inspect crossing cues individually. If one cue spans both boundaries, both its text and timing need review. Preserve deliberate simultaneous captions instead of treating every overlap as an error.

The Library of Congress describes SRT as a separate text file containing caption blocks and timecodes. Editing those blocks does not trim the video. The video excerpt and subtitle excerpt are two separate deliverables.

Clip interval: 12.000–22.000 s

ID  Original seconds  Action
40   9.000–12.000     Outside
41  11.000–14.000     Opening
42  15.000–18.000     Keep
43  20.000–23.000     Closing
44  22.000–25.000     Outside

Opening / Closing:
Review the crossing cue.

A cut through a cue is a text decision

Suppose source cue 41 says “Choose blue paper. Fold it in half.” Listening to our imagined clip establishes that only “Fold it in half” survives after the opening cut. Keep those words, with timing checked against the remaining speech. Simply clipping the original cue’s start to zero would leave a sentence about choosing paper that the viewer never hears.

At the other edge, suppose cue 43 says “Open it again. Then turn it over.” If the exported clip ends after “Open it again,” remove the absent second sentence from this clip’s subtitle copy. Do not assume you can divide a cue’s duration in proportion to its number of words: pauses and speaking rate make that unreliable. Listen to establish the actual words and their boundaries.

If a cut removes part of a word or strips away a necessary condition, reconsider the video edit. A clip containing only “use the red sheet” after omitting the spoken “do not” is misleading even if its captions match the remaining sound. Do not repair that by silently adding unheard words to the subtitles. Move the cut or obtain an approved edit that preserves the message.

W3C’s caption guidance calls for accurate speech and the non-speech audio information needed to understand the content. Check surviving sounds and speaker context too: a label that made sense in the full recording may need review when the establishing scene is gone.

Move both endpoints onto the clip clock

After deciding what remains, subtract the original clip start from each retained timestamp. A fully retained cue at 15–18 seconds becomes 3–6 seconds. Its duration stays three seconds. A crossing cue first needs its retained interval established; then that interval is moved onto the new clock.

The calculation below gives the maximum time envelope inside the clip. It does not determine word timing or approve the text. In the fictional example, listening confirms the opening phrase occupies clip seconds 0–2 and the closing phrase 8–10. In a real recording, use the audible boundaries rather than automatically extending every crossing caption to the cut.

Enter milliseconds consistently and confirm every final end is later than its start. The WebVTT specification likewise requires positive cue duration and permits overlaps. A zero-length result at a boundary is not a useful caption to retain. If the remaining phrase cannot be read comfortably, reconsider the excerpt or its caption segmentation rather than displaying words after the video ends.

S = original clip start
E = original clip end

Candidate start =
  max(cue start, S) - S
Candidate end =
  min(cue end, E) - S

Keep only if end > start.
Verify text and speech timing.

For S=12 and E=22:
41: 11–14 → 0–2 s
    Opening text reviewed
42: 15–18 → 3–6 s
43: 20–23 → 8–10 s
    Closing text reviewed

Choose a tool without losing the original

You can prepare a small excerpt in a plain-text editor by copying the selected SRT blocks into a new UTF-8 file, editing the two boundary cues and calculating the new timestamps. Use a subtitle editor for longer files so you can review the changes alongside playback. Export under a new clip-specific name and reopen it; do not overwrite the full-length track.

Aegisub documents shifting selected rows and moving both start and end times. It also documents that negative timestamps are set to zero. That makes selection essential: shifting the whole full-length file can pile discarded material at the opening. Select and resolve the excerpt before applying a backward shift, and verify the resulting export.

Polyglot can import an SRT for text and timing review and export a separate subtitle file. Its uniform timing adjustment moves all project languages together and rejects a shift that would create negative times. Prepare the excerpt as a separate file before importing it for review; do not treat that adjustment as an automatic clip extractor. This guide does not describe a Polyglot video-trimming or subtitle-cropping command.

Review each language at the new boundaries

Apply the same video interval to every language, but do not remove an equal number of words from each translation. Word order differs. A target-language cue can place a condition or important noun in a different position, so a reviewer needs to assess the retained phrase in context.

If a language has independently edited cue times, use that track’s actual intervals when selecting candidates. Do not copy the source-language timing decisions blindly. Keep the original-to-clip mapping for each delivered language, and mark boundary translations pending until someone who understands the language has reviewed them.

Deliver the clip and its matching evidence

Renumber the final SRT sequentially if needed for the destination, preserving the original-ID mapping in your notes. The illustrative export below uses the reviewed wording and times from our example. It does not include source cues 40 or 44, and it does not retain the removed sentences.

Watch the entire short video with this exact file in the intended player and with any player timing compensation reset. Check the first audible word, both cut edges, reading time, meaningful sounds and the final disappearance. Confirm the correct language and revision are selected. If you have only tested locally, leave destination acceptance pending.

Send a brief note with the clip revision, original-video interval, language, delivered filename, boundary changes and unresolved issues. A matching cue count alone cannot prove a good extraction. The useful result is a subtitle track that says what survives in this particular clip, with an explicit record of what was removed.

1
00:00:00,000 --> 00:00:02,000
Fold it in half.

2
00:00:03,000 --> 00:00:06,000
Press along the fold.

3
00:00:08,000 --> 00:00:10,000
Open it again.

Keep exploring.