Transcripts

GET /api/calls/{id}/transcript

Returns the normalised, speaker-attributed transcript for a completed call.

Response — 200 OK

{
  "schemaVersion": 2,
  "callId": "f0a1d2e3-...",
  "language": "en",
  "durationSeconds": 240,
  "cleanText": "[00:00 - 00:04] Praveen: Hi Ravi, ...\n[00:04 - 00:08] Ravi: ...",
  "speakers": [
    {
      "speakerId": "spk_1",
      "sourceSpeakerId": "SPEAKER_00",
      "label": "Praveen",
      "displayName": "Praveen",
      "role": "USER",
      "confidence": "HIGH",
      "labelSource": "GREETING_MATCH",
      "voiceMatchScore": null,
      "voiceProfileMatched": null
    },
    {
      "speakerId": "spk_2",
      "sourceSpeakerId": "SPEAKER_01",
      "label": "Ravi",
      "displayName": "Ravi",
      "role": "CONTACT",
      "confidence": "MEDIUM",
      "labelSource": "BY_ELIMINATION"
    }
  ],
  "segments": [
    {
      "id": "seg_0001",
      "speaker": "Praveen",
      "speakerLabel": "Praveen",
      "speakerDisplayName": "Praveen",
      "speakerId": "spk_1",
      "sourceSpeakerId": "SPEAKER_00",
      "role": "USER",
      "startSeconds": 0.0,
      "endSeconds": 4.2,
      "text": "Hi Ravi, I'll send the revised pricing today.",
      "alignmentConfidence": "HIGH"
    }
  ],
  "speakerResolution": {
    "strategyVersion": 2,
    "usedUserDisplayName": true,
    "usedContactName": true,
    "usedDirection": true,
    "usedPhoneFallback": false,
    "voiceEmbeddingEnabled": false,
    "voiceProfileUsed": false,
    "voiceMatchStatus": "DISABLED",
    "totalSpeakerCount": 2,
    "resolvedSpeakerCount": 2,
    "warnings": []
  },
  "pipeline": {
    "transcriptionProvider": "lemonfox",
    "transcriptionModel": "whisper-1",
    "diarizationProvider": "pyannote",
    "diarizationModel": "precision-2",
    "alignmentVersion": "1.0",
    "normalizationVersion": 3,
    "status": "COMPLETED"
  },
  "createdAt": "2026-05-29T13:00:42Z"
}

Field reference

Field	Meaning
schemaVersion	Currently 2. Bumped on breaking shape changes.
cleanText	Pretty-printed transcript suitable for direct display.
speakers[]	One entry per resolved speaker. See enum tables below.
segments[]	Time-coded segments; each refers to a speakerId.
speakerResolution	Privacy-safe telemetry block from SpeakerNameResolutionService.
pipeline	Provenance — which providers and algorithm versions produced the output.

Enums

role

Value	Meaning
USER	The authenticated user (phone owner).
CONTACT	The other party on the call.
UNKNOWN	Identity could not be determined.

confidence

Value	Meaning
HIGH	Strong evidence: e.g. a greeting addressed the other party by name.
MEDIUM	Indirect evidence (mention asymmetry, by-elimination).
LOW	Default / fallback. Always paired with POSITIONAL_FALLBACK or DIARIZATION.

labelSource

Value	Meaning
DIARIZATION	No resolution applied; raw provider speaker.
USER_PROFILE / CONTACT_METADATA	Strict text match on the supplied names.
GREETING_MATCH	Name appeared inside a greeting pattern.
NAME_MENTION	Name appeared elsewhere in the speaker's text.
BY_ELIMINATION	Other speaker was resolved; this is the remaining role.
PHONE_FALLBACK	Contact name missing; "Contact ending NNNN" was used.
POSITIONAL_FALLBACK	No evidence; direction-aware appearance-order guess.
VOICE_EMBEDDING	Voice profile matched this speaker as the user.
AMBIGUOUS	Both speakers reference both names; nothing assigned.
MANUAL	Future user-correction endpoint.

Errors

Status	code	Cause
404	call_not_found	Call missing or owned by another user.
422	call_not_completed	Status is not COMPLETED.