What the project's reference says in a window
The reference's shots (kind, framing, on-screen text, motion, palette, keyframe url), its transcript and any sound events inside a window, as rows.
Times are seconds on the SOURCE clock: from/to, or scene_id for the span the reference alignment gives that replica scene. No window = the whole source (shots cap at 200, words at 4000). Free.
Authorization
bearerAuth Typecut API key. Generate one in app.typecut.ai → Settings → API keys. Send as Authorization: Bearer tc_live_….
In: header
Path Parameters
uuidQuery Parameters
Window start, seconds on the source clock.
0 <= valueWindow end, seconds on the source clock.
A replica scene: its aligned span of the source (instead of from/to).
uuid1 adds the word timings.
Response Body
application/json
application/json
application/json
application/json
curl -X GET "https://example.com/projects/497f6eca-6276-4993-bfeb-53cbbbba6f08/reference/facts"{
"data": {
"source": {
"id": "string",
"duration_sec": 0,
"width": 0,
"height": 0,
"fps": 0,
"has_audio": true
},
"window": {
"from": 0,
"to": 0,
"scene_id": "string"
},
"shots": [
{
"i": 0,
"t0": 0,
"t1": 0,
"kind": "string",
"framing": "string",
"on_screen_text": "string",
"graphics": true,
"voice_over": true,
"motion": null,
"palette": null,
"keyframe_url": "string"
}
],
"transcript": "string",
"words": [
{
"w": "string",
"t0": 0,
"t1": 0
}
],
"events": [
{
"kind": "string",
"t0": 0,
"t1": 0,
"data": null
}
],
"coverage": {
"account": {
"property1": 0,
"property2": 0
},
"source": {
"property1": 0,
"property2": 0
}
},
"notes": [
"string"
]
}
}{
"error": {
"type": "string",
"code": "string",
"message": "string"
}
}{
"error": {
"type": "string",
"code": "string",
"message": "string"
}
}{
"error": {
"type": "string",
"code": "string",
"message": "string"
}
}A picture of the project's reference at any time, width and crop GET
Same as GET /video-analyses/{analysisId}/frames, resolved through the project's reference (404 when it has none).
Find shots of the project's reference GET
Shots whose on-screen text contains a string, or of a given kind / framing; or, with similar_to (a shot index), the nearest shots by picture (measure dino = composition and content, clip = meaning; cosine distance, smaller is closer).