Skip to main content
Descriptive alt text

​
Overview

The transcript dataset provides access to high-accuracy transcripts from regular earnings calls and unique events. Each transcript is associated with an event.
Transcripts are only available for events conducted in English.

​
How it works

Each transcript record in the API response includes a URL pointing to a JSON file hosted on our CDN. Transcripts are released in two steps:
  1. Raw transcript (type ID = 15) is published shortly after the event concludes, with paragraph breaks.
  2. Edited transcript (type ID = 22) follows once speaker identification has been completed, adding speaker names, roles, and company affiliations.

​
Chapters

Transcripts can be divided into structured segments using chapters. Use the chapters endpoint to retrieve hierarchical sections with titles and timestamps for a given transcript.

​
Speaker identification

The edited transcript (type ID = 22) maps each paragraph to a named speaker with their role and company. The JSON structure is the same as the raw transcript, with an added speaker_mapping array at the top level. Older transcripts will not be retroactively updated with speaker data. If your application filters by typeId, include both: typeIds=15,22.
It’s not always possible to identify who is speaking. In such cases, the name, role, and company fields may be null. This can happen when:
  • The speaker is not clearly identified in the audio.
  • The speaker is not part of the event’s official roster.
  • We are unable to verify the speaker’s identity or role.
During high-activity periods like earnings season, some events may be prioritized for speaker attribution based on client interest and market relevance. Speaker data may appear on certain transcripts sooner than others.

​
Data structure

Transcripts are structured as a hierarchy: the top-level transcript object contains the full text and an array of paragraphs, each paragraph contains sentences, and each sentence contains words. Every level includes start/end timestamps in seconds and a zero-based speaker index. 
{
  "version": "1.0",
  "event_id": 123456,
  "company_id": 123,
  "transcript": {
    "text": "This is the full transcript text",
    "number_of_speakers": 3,
    "paragraphs": [
      {
        "text": "This is the paragraph text",
        "start": 0,
        "end": 10,
        "speaker": 0,
        "sentences": [
          {
            "text": "This is the sentence text",
            "start": 0,
            "end": 5,
            "words": [
              {
                "word": "This",
                "punctuated_word": "This",
                "start": 0,
                "end": 5,
                "confidence": 0.9
              }
            ]
          }
        ]
      }
    ]
  }
}

​
How to access this data

REST API

Query audio files using company and event filters for full control.

Webhooks

Subscribe to webhooks for real-time updates.

Snowflake

Query the transcripts view directly using SQL.