# Decide whether a subtitle overlap needs fixing

Find overlapping cues, including a long subtitle spanning several shorter ones, and choose a repair using the video.

Source: https://subtitleworkbench.com/guides/subtitle-overlap

## Measure the shared interval first

If cue A ends at 4.000 seconds and cue B begins at 3.750 seconds, the intervals share a quarter of a second. Whether that needs correction depends on how the captions are used. Simultaneous speakers or independently positioned text can be intentional.

WebVTT explicitly permits overlapping cue timings. SRT is less formally specified, and players differ in how they present simultaneous cues. Subtitle Workbench therefore reports an overlap as a review finding rather than a universal format error.

```text
Cue A: 00:00:01,000 --> 00:00:04,000
Cue B: 00:00:03,750 --> 00:00:06,500
Shared interval: 3.750 through 4.000 seconds
Overlap: 0.250 seconds
```

## Compare the two boundary changes

Ending A at 3.750 seconds removes the overlap by shortening A. Starting B at 4.000 seconds removes it by delaying B. The first change reduces reading time for the earlier subtitle; the second moves the later subtitle’s start.

A fixed shift applied to every cue preserves the distance between timestamps, so it cannot remove an overlap. Renumbering also has no effect on timing. A cleaner-looking export is not necessarily a correction to the viewing experience.

- For an accidental overlap, use the audio and source project to choose the right boundary. Then check cue duration, reading speed, and any minimum-gap requirement from the client. An overlap warning does not supply that required gap.
- For an intentional overlap, test the target player. A format can represent simultaneous cues without every player laying them out as expected. Look for conflicting lines or text that obscures important visual content.

| Change | What it preserves | What it changes |
| --- | --- | --- |
| End A at 3.750 | B’s start | A’s reading time |
| Start B at 4.000 | A’s end | B’s start time |

## Try a repair that changes the reading speed

Paste this invented practice file into the checker. Keep the default review settings and leave Time shift, seconds at zero. The first cue has 57 counted characters, including spaces and punctuation but excluding its line break. It lasts 3 seconds, so its reading speed is 19.0 CPS. The only finding should be the 0.250-second overlap on block 2.

```text
1
00:00:01,000 --> 00:00:04,000
Please keep the doors closed until
the train stops moving.

2
00:00:03,750 --> 00:00:06,500
I will wait.

```

- Select the overlap finding for block 2. In Updated block, change only its start from 00:00:03,750 to 00:00:04,000. Compare Before this edit, then select Apply block. Both sentences stay intact and the finding disappears.
- Select Undo edit to restore the overlap. To try the other repair, change the first cue’s end from 00:00:04,000 to 00:00:03,750 in SRT source. Keep the second cue at its original times. The overlap disappears, but 57 ÷ 2.75 is about 20.7 CPS, above the default 20 CPS threshold.
- The table compares each repair with the original; do not apply both changes together. With custom review settings, the findings may differ.

| Version | First cue duration / CPS | Second cue duration | Default findings |
| --- | --- | --- | --- |
| Original | 3.000 s / 19.0 CPS | 2.750 s | Overlap on block 2 |
| End first cue at 3.750 | 2.750 s / 20.7 CPS | 2.750 s | Reading speed on block 1 |
| Start second cue at 4.000 | 3.000 s / 19.0 CPS | 2.500 s | None |

## Choose the boundary with the video open

The version with no findings still needs playback review. Delaying the second cue can make its words appear late; shortening the first can remove text before its speech ends. These invented timestamps cannot tell you which edit matches your video.

Both repairs leave a zero-second gap. If the delivery specification requires a positive gap, there is another timing decision to make. Keep checking the affected cue durations and reading speeds after that change. A global time shift moves both boundaries equally and leaves the original overlap unchanged.

## Look beyond the next cue

A long cue can stay on screen across several shorter cues. In this invented example, block 1 runs from 1 to 10 seconds, block 2 from 2 to 3, and block 3 from 4 to 5. Blocks 2 and 3 do not overlap each other, but each overlaps block 1 for one second. Paste all three blocks: the checker should show an overlap on both block 2 and block 3, naming block 1 in each finding.

Each overlap finding names one intersecting block and the duration shared by that pair. The warning goes on the cue that starts later; if their start times match, it goes on the later block in the file. A cue may overlap more blocks than the single reference shown. Resolve or retain the overlap deliberately, then check again.

Timeline checking also works when the file is out of order, while a separate finding flags a start time that moves backward. References always use the block positions in your source, even if the printed cue numbers are wrong. Checking does not sort, shorten or merge the exported cues. Only readable cues can be compared; format errors still block export.

```text
1
00:00:01,000 --> 00:00:10,000
The station remains closed.

2
00:00:02,000 --> 00:00:03,000
Wait here.

3
00:00:04,000 --> 00:00:05,000
I understand.

```

## Sources

- [W3C WebVTT specification: cue timings and overlap](https://www.w3.org/TR/webvtt1/)
- [Library of Congress: SubRip format](https://www.loc.gov/preservation/digital/formats/fdd/fdd000569.shtml)

[Try the SRT checker](https://subtitleworkbench.com/) | [Method and limitations](https://subtitleworkbench.com/about)