Subtitles

Automatic speech recognition

Understand how the transcript is produced and what to review after automatic recognition.

Reading time: 4 min

How transcription works

After upload, the service extracts the audio, recognizes speech, and calculates word timings. The transcript is then divided into short cues shown in the editor and video preview.

What affects accuracy

  • Voice clarity and volume.
  • Background music and noise.
  • Several people speaking at once.
  • Unusual names, abbreviations, and specialist terms.
  • Strong echo, microphone clipping, and low audio bitrate.

What you should always review

  1. Listen to the beginning and confirm that the detected text matches the speech.
  2. Check names, brands, addresses, numbers, and dates.
  3. Review sections with music, pauses, or speaker changes.
  4. Confirm that cues appear and disappear at the right time.