> ## Documentation Index
> Fetch the complete documentation index at: https://benzinga.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Overview

> Real-time and historical conference call transcripts, audio, summaries, and speaker data.

## Introduction

The Conference Call Transcripts API gives you programmatic access to corporate
earnings and conference calls — the **full text transcript, the audio, an AI-generated
summary, and structured speaker data** — for both **live calls as they happen** and
**completed calls** from the archive.

Every call is transcribed in real time. From the moment a call goes live you can:

* **Read it as it's spoken** — sentence-by-sentence transcript chunks pushed over WebSocket.
* **Listen to it live** — an HLS audio stream that becomes available at the call's start time.
* **Fetch the finished product** — the complete transcript, MP3 audio, AI summary, sentiment,
  and identified participants once the call ends.

There are two ways to consume the data, and most integrations use both together:

<CardGroup cols={2}>
  <Card title="WebSocket — live" icon="bolt">
    Subscribe to tickers and receive live transcript sentences and the live audio link
    the moment a call starts. Push-based, no polling.
  </Card>

  <Card title="REST — on demand" icon="server">
    Fetch any call (live or completed) by ID or list them with filters. Returns the full
    transcript, audio URLs, summary, sentiment, and participants.
  </Card>
</CardGroup>

## The call lifecycle

A single call moves through the same lifecycle whether you watch it live or pull it later.
The data model is **the same object throughout** — nothing appears and then disappears; the
completed call is simply the live call with more fields filled in.

<Steps>
  <Step title="Call starts">
    The transcript stream begins emitting sentence chunks over WebSocket, and the **live HLS
    audio stream becomes available** at the call's start time. Over REST the call appears with
    `status: STARTED` / `IN_PROGRESS`.
  </Step>

  <Step title="Call is live">
    Transcript sentences stream in continuously. The `recordings` object carries a single
    **HLS** format (`application/vnd.apple.mpegurl`, `720p`) — the live, growing audio stream.
  </Step>

  <Step title="Call completes">
    The transcript is finalized and the audio playlist is sealed. An **MP3** format is added
    alongside the HLS one, so a completed call's `recordings` has two formats (HLS + MP3).
  </Step>

  <Step title="Call is enriched">
    Post-processing adds the **AI summary**, **per-segment and overall sentiment**, **speaker
    labels and identified participants**, and a downloadable **PDF** transcript.
  </Step>
</Steps>

## What's in a call

A fetched call is a single JSON object that bundles everything about the call:

| Field                                          | Description                                                                                                        |
| :--------------------------------------------- | :----------------------------------------------------------------------------------------------------------------- |
| `call_id`                                      | Unique identifier for the call — the key you use everywhere.                                                       |
| `symbol`, `name`, `exchange`                   | The company the call belongs to.                                                                                   |
| `start_time`, `end_time`, `duration`, `status` | Call timing and lifecycle state.                                                                                   |
| `transcripts[]`                                | The transcript text, broken into `segments` with `speaker`, timestamps, `confidence`, and per-segment `sentiment`. |
| `recordings[].formats[]`                       | Audio files — HLS (live + recorded) and MP3 (recorded). Each has a `file_link`, `content_type`, and `quality`.     |
| `summary`                                      | AI-generated summary of the call plus an overall `sentiment` score.                                                |
| `participants[]`                               | Identified speakers with their `name`, `role`, `organization`, and mention count.                                  |

<Note>
  Audio URLs are only included when you request them — pass `audio=true` on the
  [Fetch Call by ID](/api-reference/delivery-api/calls/fetch-call-by-id) endpoint. Without it,
  the `recordings` block is omitted from the response.
</Note>

## Live audio (HLS)

Live and recorded audio are served the same way: as a signed, time-limited **HLS**
(`.m3u8`) stream from Benzinga's CDN. The HLS stream is available from the call's start
time and simply stops growing when the call ends; the MP3 is added on completion. There are
two ways to get the audio URL.

<Tabs>
  <Tab title="WebSocket (live)">
    After you subscribe to a ticker, the **first message for that ticker is an `audio`
    event** carrying the live HLS URL — you don't poll for it, it's pushed to you. Every
    message after that is a transcript sentence.

    ```jsonc theme={null}
    // 1. Connect: wss://api.benzinga.com/api/v1/transcripts/stream?token=YOUR_API_KEY
    // → {"message":"Connected to the server","uid":"..."}

    // 2. Subscribe
    {"action":"subscribe","ticker":"MSM"}
    // → "Subscribed to MSM"

    // 3. The audio event — carries the live HLS link
    {
      "action":"audio",
      "symbol":"MSM",
      "call_id":"6a390146b856ac00017f8fe5",
      "hls_url":"https://api.benzinga.com/api/v1/transcripts/calls/6a390146b856ac00017f8fe5/hls?Key-Pair-Id=...&Policy=...&Signature=..."
    }

    // 4. Then transcript sentences stream in
    {"call_id":"6a390146b856ac00017f8fe5","symbol":"MSM","sentence":"...","status":"IN_PROGRESS", ...}
    ```

    Subscribe to all live calls with `{"action":"subscribe","ticker":"*"}` — you'll receive
    an `audio` event for every call that currently has a live stream.
  </Tab>

  <Tab title="REST (on demand)">
    Fetch the call with `audio=true`. The audio URL lives in the `recordings` block:

    ```bash theme={null}
    GET https://api.benzinga.com/api/v1/transcripts/calls/{call_id}?token=YOUR_API_KEY&audio=true
    ```

    ```jsonc theme={null}
    // Audio lives at recordings[0].formats[]
    "recordings": [
      {
        "formats": [
          {
            "file_link": "https://.../hls/index.m3u8?<signed>",
            "content_type": "application/vnd.apple.mpegurl",
            "quality": "720p"        // live + recorded
          },
          {
            "file_link": "https://.../mp3/audio.mp3?<signed>",
            "content_type": "audio/mpeg",
            "quality": "192kbps"      // added when the call completes
          }
        ]
      }
    ]
    ```

    <Warning>
      Without `audio=true` the `recordings` block is omitted entirely — no audio URLs are
      returned. This is the most common reason audio appears "missing" from a response.
    </Warning>
  </Tab>
</Tabs>

### Playing the stream

The audio URL is a signed, time-limited HLS stream. Open it in any HLS-capable player:

* **Safari / Chrome** — play HLS natively; paste the URL straight into the address bar.
* **VLC** — Media → Open Network Stream → paste the URL.
* **In-browser** — use [hls.js](https://github.com/video-dev/hls.js).

## Available endpoints

<CardGroup cols={2}>
  <Card title="Fetch All Calls" icon="list" href="/api-reference/delivery-api/calls/fetch-all-calls">
    List calls with filtering and pagination — by ticker, date range, and status.
  </Card>

  <Card title="Fetch Call by ID" icon="file-lines" href="/api-reference/delivery-api/calls/fetch-call-by-id">
    Retrieve one call in full: transcript, audio, summary, sentiment, and participants.
  </Card>

  <Card title="Fetch Summaries" icon="align-left" href="/api-reference/delivery-api/summaries/fetch-summaries">
    List AI-generated call summaries with pagination.
  </Card>

  <Card title="Fetch Summary by Call ID" icon="wand-magic-sparkles" href="/api-reference/delivery-api/summaries/fetch-summary-by-call-id">
    Retrieve the summary and overall sentiment for a specific call.
  </Card>

  <Card title="Transcripts Stream (WebSocket)" icon="bolt" href="/ws-reference/data-websocket/get-transcripts-stream">
    Subscribe to live transcript sentences and live audio links as calls happen.
  </Card>
</CardGroup>