Reference

Transcript JSON Format

The export and import feature uses a single JSON document per interview, format_version 1.

{
  "format": "oral-archivist-transcript",
  "format_version": 1,
  "language": "en",
  "duration_ms": 5292000,
  "paragraphs": [
    {
      "start_ms": 0,
      "end_ms": 14480,
      "speaker": "Eleanor Marsh",
      "text": "We would walk to the [[fn:school1]]schoolhouse[[/fn]] daily."
    }
  ],
  "resolved_flags": [],
  "chapters": [
    {"start_ms": 0, "end_ms": 860000, "title": "Childhood on the farm",
     "topics": ["Childhood", "Rural schools"]}
  ],
  "footnotes": [
    {"anchor_key": "school1", "body": "The one-room school stood until 1961.",
     "url": "https://example.org/schools", "resource_id": null,
     "resource_type": null, "is_public": true}
  ]
}

Fields

Field Required Description
format yes Always oral-archivist-transcript.
format_version yes Currently 1.
language no ISO 639-1 code.
duration_ms no Total duration in milliseconds.
paragraphs yes Ordered list of paragraphs.
paragraphs[].start_ms, end_ms yes Paragraph time range in milliseconds.
paragraphs[].speaker no Speaker name. Speakers are matched and created by name on import.
paragraphs[].text yes Paragraph text, optionally containing footnote anchor tokens.
resolved_flags no Raw word indexes whose low-confidence flags have been marked reviewed.
chapters no Chapters with start_ms, end_ms, title and a topics list of labels.
footnotes no One entry per anchor key used in the text.

Footnote anchors

Anchors are written inline as [[fn:KEY]]words[[/fn]]. Keys match [A-Za-z0-9_-]{1,40} and must be unique within the interview, with exactly one footnotes entry per key. Each footnote needs at least one of body, url, or a resource_id with resource_type of items or media. is_public defaults to true.

Import behaviour

  • Paragraphs, speakers, chapters and footnotes are replaced wholesale.
  • The interview always returns to the review queue.
  • Raw word-level timestamps from transcription are never modified. Word times used by search-to-moment are recomputed at publish.
  • Footnotes that reference resources missing on the importing site are kept with the resource link removed, and a warning is reported.
Log in for Support