Caption Cue Window Text Extractor
Extract ID/text observations from cues with positive half-open overlap of an explicit media window; not a transcript alignment model.
Description
Extract ID/text observations from cues with positive half-open overlap of an explicit media window; not a transcript alignment model.
Window-based caption extraction uses positive temporal overlap, so a cue touching only a boundary is excluded.
When to use this tool
Extract ID/text observations from cues with positive half-open overlap of an explicit media window; not a transcript alignment model.
- Prepare the supplied cuesJson, start, end fields for this operation.
- Compare the complete example with the representation used by your own input before carrying the result into another step.
How to use the tool and how it works
- Enter cuesJson, start, end. Use the exact text syntax, units and index conventions explained below.
- Run the tool. If input validation fails, correct the supplied fields before interpreting any result.
- Read result.texts. Interpret false, null and empty lists using the distinctions below; they can be valid outputs.
Require nonnegative ordered millisecond start/end bounds. Keep cues where max(cue start,window start) is strictly below min(cue end,window end), returning ID/text records.1
| Input field | Meaning and representation | Accepted input |
|---|---|---|
| cuesJson | JSON text, entered as one text field. The decoded structure is illustrated below; it is not a separate file. | Text |
| start | start. | Integer from 0 to 1e12 inclusive |
| end | end. | Integer from start to 1e12 inclusive |
For the JSON text fields, the following decoded structures show the example records before they are serialized into input strings. Keep field names, list nesting and value types as shown.
[
{
"field": "cuesJson",
"decoded": [
{
"id": "a",
"start": 0,
"end": 4,
"text": "hi"
},
{
"id": "b",
"start": 3,
"end": 6,
"text": "hi"
}
]
}
]| Output field | Interpretation |
|---|---|
| result | Object containing the named outputs below. |
| result.texts | The texts list in the ordering and representation described above. |
{
"input": {
"cuesJson": "[{\"id\":\"a\",\"start\":0,\"end\":4,\"text\":\"hi\"},{\"id\":\"b\",\"start\":3,\"end\":6,\"text\":\"hi\"}]",
"start": 4,
"end": 5
},
"output": {
"result": {
"texts": [
{
"id": "b",
"text": "hi"
}
]
}
}
}Limitations and assumptions
- No cue timing is clipped and no text is merged. A zero-width window or zero-duration cue does not contribute.
- Input cue records contain string id/text and integer millisecond start/end from zero to one trillion, with end at least start. At most 128 cue records are accepted; IDs need not be unique.
- Every declared top-level field is required and unexpected fields are rejected. Numeric inputs must be finite; ordinary text fields are bounded to 8,192 UTF-16 code units, with tighter operation-specific limits where stated.
- Lists are limited to 128 entries. JSON text must meet bounded size and depth checks, and numeric outputs beyond the safe finite magnitude are rejected. Exact large integers use decimal strings.
- Indexes are zero-based and spans have an exclusive end unless explicitly stated otherwise. A valid false, null or empty-list result is distinct from malformed input.
Alternative or Complementary methods
Use a subtitle-format parser or media review for markup, file syntax and synchronization. Preserve the integer-millisecond timeline and the stated overlap or clipping convention when composing these operations.
References
Similar or alternative tools
- Caption Cue Clip to Media Window
Clip caption intervals to a half-open media window and discard empty intersections without altering IDs or text.
- Caption Empty Text Audit
Identify caption cues whose text is empty after JavaScript whitespace trimming; does not interpret markup or infer speech.