Output

Response Format

The API returns JSON objects containing transcription results in real-time.

Success Response

Complete Segment

{
  "uid": "client-identifier",
  "segments": [
    {
      "start": "0.000",
      "end": "2.500",
      "text": "Hello world, this is a test.",
      "completed": true,
      "segment_id": "seg_001",
      "rev": 1
    }
  ],
  "language": "en",
  "language_probability": 0.95
}

Partial Segment

{
  "uid": "client-identifier",
  "segments": [
    {
      "start": "0.000",
      "end": "1.200",
      "text": "Hello",
      "completed": false,
      "segment_id": "seg_001",
      "rev": 1
    }
  ]
}

Response Fields

Field	Type	Description
uid	string	Client identifier for the session
segments	array	Array of transcription segments
segments[].start	string	Start time in seconds (3 decimals)
segments[].end	string	End time in seconds (3 decimals)
segments[].text	string	Transcribed text content
segments[].completed	boolean	true if final, false if partial
segments[].segment_id	string	Unique segment identifier
segments[].rev	integer	Revision number
language	string	Detected or specified language
language_probability	float	Confidence score (0–1)

Understanding Segment Updates

You'll receive transcription results in real-time as the audio is processed:

Partial segments (completed: false) as speech is being processed
Complete segments (completed: true) when speech boundaries are detected
Updated segments with the same segment_id but higher rev number if corrections are made

Example flow:

// First update (partial)
{ "segment_id": "seg_001", "text": "Hello", "completed": false, "rev": 1 }

// Second update (still partial, more text)
{ "segment_id": "seg_001", "text": "Hello world", "completed": false, "rev": 2 }

// Final update (complete)
{ "segment_id": "seg_001", "text": "Hello world", "completed": true, "rev": 3 }

Working with Results

Display only completed segments:

def on_message(ws, message):
result = json.loads(message)

if result.get('segments'):
    for segment in result['segments']:
        if segment['completed']:
            print(f"[{segment['start']}s - {segment['end']}s]: {segment['text']}")

Track segment revisions:

const segments = new Map();

function handleResult(result) {
  if (result.segments) {
    result.segments.forEach(segment => {
      const existing = segments.get(segment.segment_id);

      if (!existing || segment.rev > existing.rev) {
        segments.set(segment.segment_id, segment);
      }
    });
  }
}

Check language detection confidence:

if result.get('language_probability'):
prob = result['language_probability']
lang = result['language']

if prob > 0.9:
    print(f"Confident detection: {lang}")
elif prob > 0.7:
    print(f"Moderate confidence: {lang} ({prob:.2%})")
else:
    print(f"Low confidence: {lang} ({prob:.2%})")