Recognition & accuracy

Biggest wins — audio & setup

The setup choices that affect accuracy most, before any tuning.

Audio quality comes first

For live events, input quality usually matters more than any software setting. A clean, consistent audio signal in the correct language is the single biggest driver of accurate captions.

  • Microphone source: use a dedicated pulpit, lapel, or line/USB microphone rather than the laptop's built-in mic, which picks up room noise and reverb.
  • Levels: aim for a clear, consistent signal — too quiet causes missed words, too loud causes distortion.
  • Speaker pace: a natural pace with slight pauses between thoughts is recognized more reliably than fast, run-on speech.
  • Correct source language: make sure the session's spoken language matches what is actually being said.
  • Fewer target languages: translate only the languages you need.
  • Network: use a stable connection and avoid guest networks that block real-time connections.

Rehearse with the real microphone on the real network before going live. This catches the issues that matter most.