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

# Working with Transcripts

> Understanding transcript formats, word timestamps, and speaker labels

## Transcript Formats

Transcord provides transcripts in three formats:

| Format   | Use Case                                                          |
| -------- | ----------------------------------------------------------------- |
| **JSON** | Programmatic access with word-level timestamps and speaker labels |
| **Text** | Simple plain text for reading or processing                       |
| **SRT**  | Subtitle format for video editing software                        |

## JSON Format (Default)

The JSON format includes the full transcript text plus word-level timing and speaker information.

```bash theme={null}
curl -X GET "https://app.transcord.app/api/v1/recordings/RECORDING_ID/transcript" \
  -H "Authorization: Bearer tr_live_your_api_key"
```

```json Response theme={null}
{
  "recordingId": "clx1234567890",
  "duration": 342,
  "toNumber": "+14155551234",
  "fromNumber": "+14155555678",
  "createdAt": "2026-01-15T10:30:00.000Z",
  "transcript": {
    "text": "Hello, this is John calling about the interview...",
    "wordCount": 847,
    "words": [
      {
        "word": "Hello,",
        "start": 0.5,
        "end": 0.9,
        "speaker": "you"
      },
      {
        "word": "this",
        "start": 1.0,
        "end": 1.2,
        "speaker": "you"
      },
      {
        "word": "Hi",
        "start": 2.1,
        "end": 2.3,
        "speaker": "them"
      }
    ]
  }
}
```

### Word Object

| Field     | Type   | Description                                       |
| --------- | ------ | ------------------------------------------------- |
| `word`    | string | The transcribed word including punctuation        |
| `start`   | number | Start time in seconds from beginning of recording |
| `end`     | number | End time in seconds                               |
| `speaker` | string | `"you"` (caller) or `"them"` (recipient)          |

## Plain Text Format

Get just the transcript text without timestamps:

```bash theme={null}
curl -X GET "https://app.transcord.app/api/v1/recordings/RECORDING_ID/transcript?format=text" \
  -H "Authorization: Bearer tr_live_your_api_key" \
  -o transcript.txt
```

```text Response theme={null}
Hello, this is John calling about the interview. I wanted to follow up on our conversation from last week...
```

## SRT Subtitle Format

Get the transcript as SRT subtitles for video editing:

```bash theme={null}
curl -X GET "https://app.transcord.app/api/v1/recordings/RECORDING_ID/transcript?format=srt" \
  -H "Authorization: Bearer tr_live_your_api_key" \
  -o transcript.srt
```

```srt Response theme={null}
1
00:00:00,500 --> 00:00:05,200
Hello, this is John calling about the interview.

2
00:00:05,800 --> 00:00:10,100
I wanted to follow up on our conversation from last week.

3
00:00:11,000 --> 00:00:15,500
Hi John, thanks for calling back.
```

## Working with Timestamps

### Jump to a specific moment

Use the `start` timestamp to link directly to a moment in the audio:

```javascript theme={null}
// Find when a specific phrase was said
const words = transcript.words;
const targetPhrase = "the contract";

const match = words.find(w =>
  w.word.toLowerCase().includes(targetPhrase.split(' ')[0])
);

if (match) {
  console.log(`"${targetPhrase}" was said at ${match.start} seconds`);
  // Use this timestamp with the audio player
}
```

### Extract quotes with timestamps

```javascript theme={null}
// Extract a quote between two timestamps
function extractQuote(words, startTime, endTime) {
  return words
    .filter(w => w.start >= startTime && w.end <= endTime)
    .map(w => w.word)
    .join(' ');
}

const quote = extractQuote(transcript.words, 45.0, 52.5);
console.log(`Quote: "${quote}"`);
```

### Group by speaker

```javascript theme={null}
// Group consecutive words by speaker
function groupBySpeaker(words) {
  const groups = [];
  let currentGroup = null;

  for (const word of words) {
    if (!currentGroup || currentGroup.speaker !== word.speaker) {
      currentGroup = { speaker: word.speaker, text: '', start: word.start };
      groups.push(currentGroup);
    }
    currentGroup.text += word.word + ' ';
    currentGroup.end = word.end;
  }

  return groups;
}
```

## Dual-Channel Recording

Transcord records calls in dual-channel format, meaning your voice and the other party's voice are recorded on separate audio channels. This enables:

* **Accurate speaker identification** - We know exactly who said what
* **Better transcription quality** - No cross-talk confusion
* **Clean audio separation** - Each speaker's audio is crystal clear

The `speaker` field in the transcript reflects this:

* `"you"` - Words spoken by you (the Transcord user)
* `"them"` - Words spoken by the other party

## Transcript Not Ready

If you request a transcript before processing is complete, you'll receive a `202 Accepted` response:

```json theme={null}
{
  "error": "Transcript not ready",
  "status": "transcribing"
}
```

Poll again in a few seconds. Status values:

* `processing` - Call just ended, preparing recording
* `transcribing` - Audio is being transcribed
* `completed` - Transcript is ready
* `failed` - Transcription failed (rare)