Two different workflows
Live recognition starts with the user speaking and ends with text in the active app. Transcription works from an existing file. They should not be judged by the same metric: dictation needs low friction, while an interview needs completeness and useful editing.
How to use it on Mac
Echo starts short dictation with Command-D and copies the finished text to the clipboard. Command-R resets the current capture. The app runs on macOS 14 or later and supports Intel and Apple Silicon.
For cloud processing, the user chooses OpenAI or Grok and adds a personal API key. The key is stored in macOS Keychain, and audio goes directly to the selected provider.
Improve recognition quality
- speak close to the microphone;
- avoid music and constant background noise;
- slow down slightly for names and numbers;
- verify important passages against the source;
- treat automatic transcription as a draft, not final editing.
Compare dictation on Mac with voice input on Mac, and use audio-to-text transcription for longer recordings.
Conclusion
Speech recognition on Mac is most useful when live input and file transcription are treated as separate jobs, with a clear launch action, transparent audio path, and time for final review.