Subtitle timing: a fixed offset or a growing drift?
When a subtitle appears before its speaker talks, the cause may be a simple offset or a different video version. Compare an early and a late line before shifting the whole subtitle file.
A subtitle file combines text with display times
A cue connects its start and end times to the words displayed. SRT and VTT are formats for this purpose; converting between them neither translates the language nor extracts speech from audio.
WebVTT uses UTF-8 and supports more than basic text cues. A .vtt extension does not mean every feature can be transferred unchanged to SRT.
References: W3C — WebVTT structure, UTF-8 and cue timings · Library of Congress — SubRip subtitle format
An offset fixes a constant timing error
If every line appears approximately the same amount too early, a positive offset delays its start and end together. The display duration stays the same, and the wording need not change.
If the mismatch is small near the beginning and larger later, one fixed shift cannot align both points. Check that the video and subtitle versions match; gradual drift or detailed retiming requires an editor designed for that task.
Know what the converted copy can retain
ToolRoof converts basic text cues between SRT and VTT and applies a constant offset. It supports multiline text, but rejects formatting tags, text entities, VTT positioning settings and cue identifiers rather than silently dropping them.
It does not burn subtitles into a video or transcribe audio. Keep the source, open the result with the intended video, and check an early and a late line. An offset must not push a cue’s start before time zero.
Example: delaying a cue by 750 milliseconds
A cue running from 00:00:10,500 to 00:00:12,000 lasts 1500 milliseconds. Adding 750 milliseconds produces VTT times of 00:00:11.250 to 00:00:12.750; its display duration is unchanged.
If the early mismatch is 750 milliseconds but the late mismatch is 3000 milliseconds, adding 750 leaves a 2250-millisecond error near the end. That is evidence that a single fixed offset is not enough.