Edit and publish¶
Utterance text preserves spaces, punctuation, Unicode, and explicit separators. Word timing is stored separately.
Work with utterances¶
Open a transcript from /admin/content/transcriptions. Its tabs provide metadata, the utterance editor, speaker mapping, revision history, and exports.
The editor supports correcting text, appending utterances, splitting and merging, removing, and reordering. Additional configured fields use their ordinary Drupal form displays. Editing is paged, with 25 utterances per page by default and a configurable maximum of 100.
Corrections preserve unaffected token identities and times. A one-token replacement can retain the old interval as stale; added or replaced phrases can become untimed. Review the affected intervals before confirming their alignment.
Assign speakers¶
Map a provider’s speaker cluster to an allowed Drupal entity, or assign an individual utterance. An individual override survives later cluster remapping. When no accessible entity is mapped, the reader can show the raw speaker label.
The site’s Transcribe settings control which entity types and bundles are eligible. Speaker access remains part of the transcript’s access rules.
Save and review revisions¶
A transcription revision owns an ordered list of utterance revisions. Saving edits coordinates both in one database transaction. Published revisions retain their existing graph when a new draft is saved.
The editor rejects stale submissions. Reload and reconcile your edits with the newer revision. Restoring a historical transcript creates a new parent revision and preserves intervening history.
See Access and revisions for the full access and save contract.
Publish and play¶
Use Drupal’s publication controls after reviewing text, speakers, and timing. The reader inherits the site theme and supports native audio/video plus explicit YouTube loading.
Timed words are native buttons: activate one to seek the source. Playback highlights suitable measured or retained timing. Untimed, stale, and estimated text remains readable. Follow playback scrolling is opt-in and respects reduced motion.
Export¶
Export the selected revision as exact TXT, named JSON, SRT, or WebVTT. Caption formats use real utterance ranges and skip unknown ones. Access to the transcript, its graph, and source is checked before export.
Public services and exchange format documents the data and export API.