One subtitle track can contain several active cues at the same moment. If two lines remain after you have ruled out burned-in text and a second selected track, inspect the cue intervals inside the WebVTT file. An overlap can be valid, especially for simultaneous speech; the task is to decide whether this particular concurrency is intentional.
Inspect active intervals rather than track count
The W3C WebVTT specification permits overlapping cues and orders cues by start time. A valid file is therefore not necessarily a single-caption-at-a-time file. Rendering and placement can vary with the player, cue settings and available space, so two active cues need not appear exactly where an editor preview placed them.
Work on a copy and find the moment with the unwanted display. List every cue whose start is at or before that moment and whose end is later. Use this half-open interval convention for the diagnostic ledger so a cue ending exactly when another begins is not counted as a genuine overlap. Then read the associated words against the actual audio.
Classify the reason before editing
In a fictional example, cue A runs from 5.000 to 8.000 seconds and cue B from 7.000 to 9.000. Both are active for one second. That arithmetic explains concurrency, but it does not tell you which text should be removed.
| Content relationship | Appropriate review |
|---|---|
| Same sentence duplicated by an import | Remove the confirmed duplicate |
| New line starts while old speech has ended | Correct the old cue's end against audio |
| Two people genuinely speak together | Preserve the necessary content with clear attribution |
| Intentional bilingual presentation | Review the chosen combined layout and timing |
Do not shorten every cue to the next start as a blind cleanup. That can erase a quiet speaker's continuing sentence or remove an accessibility sound cue that remains relevant. Likewise, merging all concurrent words into one long line may create a reading problem even though only one cue remains.
Verify the repaired passage in the real player
Play from before the first cue through the end of both. Toggle the single track to confirm that you are still examining that file, then test the intended viewport and caption preferences. Inspect whether concurrent cues cover faces or exchange vertical order in a confusing way. This is an authored test plan, not a browser conformance measurement.
Record which overlap was accidental and which was deliberately retained. The acceptance condition is that the audience receives the intended concurrent information without an unexplained duplicate. It is not a global count of zero overlapping timestamps.
For an Imagild film with externally authored captions, bind the reviewed WebVTT to the exact final edit. If a later recut shifts only one group of cues, repeat the interval review around those edits. A single file and valid syntax establish neither the correct editorial timing nor a universally identical visual layout.