Skip to main content
View as Markdown

Transcripts - Content Management

Transcripts​

Transcript vs Caption​

  • Caption (KalturaCaptionAsset): Timed text segments with start/end timestamps. Displayed as subtitles during playback.
  • Transcript (TranscriptAsset): Full text document derived from the audio track. Extends TextualAttachmentAsset (separate from CaptionAsset). Linked to caption assets via the associatedTranscriptIds field.

In practice, REACH-generated captions serve as both: the timed segments display as subtitles, and the full text is available for search and download.

Machine Transcription via REACH​

Order automatic speech recognition (ASR) via REACH:

curl -X POST "$KALTURA_SERVICE_URL/service/reach_entryVendorTask/action/add" \
  -d "ks=$KALTURA_KS" \
  -d "format=1" \
  -d "entryVendorTask[objectType]=KalturaEntryVendorTask" \
  -d "entryVendorTask[entryId]=$KALTURA_ENTRY_ID" \
  -d "entryVendorTask[reachProfileId]=$REACH_PROFILE_ID" \
  -d "entryVendorTask[catalogItemId]=$ASR_CATALOG_ITEM_ID"
  • serviceFeature: 1 (CAPTIONS)
  • serviceType: 2 (MACHINE)
  • Accuracy: ~83-87%
  • Turnaround: Near-instant (minutes)

See REACH API for full REACH task configuration.

Human Transcription via REACH​

Order professional human transcription:

  • serviceFeature: 1 (CAPTIONS)
  • serviceType: 1 (HUMAN)
  • Accuracy: 99%+
  • Turnaround: 3-72 hours depending on vendor and turnaround tier
  • Vendors: Professional human transcription vendors available through the REACH marketplace

Transcript Alignment via REACH​

Upload existing text and let REACH align it to the audio timeline:

  • serviceFeature: 1 (CAPTIONS)
  • Provide the text via the task's inputMetadata field
  • REACH syncs text to audio, creating timed caption segments

Checking Task Status​

curl -X POST "$KALTURA_SERVICE_URL/service/reach_entryVendorTask/action/getJobs" \
  -d "ks=$KALTURA_KS" \
  -d "format=1" \
  -d "filter[objectType]=KalturaEntryVendorTaskFilter" \
  -d "filter[entryIdEqual]=$KALTURA_ENTRY_ID"

Task status flow:

ValueNameDescription
8PENDING_ENTRY_READYWaiting for entry to reach READY status
1PENDINGQueued for processing
3PROCESSINGBeing transcribed
2READYComplete
6ERRORProcessing failed

With moderation enabled: PENDING_MODERATION (4) requires explicit approve/reject before delivery.

Output​

When a REACH task completes, it creates a KalturaCaptionAsset attached to the entry automatically. The outputObjectId on the completed task references the created caption asset ID. The caption is set as default if no default exists.

Custom Vocabulary​

REACH supports custom dictionaries for domain-specific terminology. Configure vocabulary lists in the REACH profile to improve accuracy for specialized content (medical, legal, technical terms). See REACH API for dictionary management.