Lagos Apple Speech (Level 1)
Apple Speech is Lagos’s default transcription engine. It uses macOS’s built-in Speech Recognition framework — the same technology powering Siri dictation. No downloads, no API keys, no configuration.
How It Works
Section titled “How It Works”When you hold your shortcut and speak, Lagos captures audio via AVAudioEngine and streams it to the macOS Speech Recognition API. Transcription happens on-device, with results returned in near real-time.
Accuracy
Section titled “Accuracy”Apple Speech has approximately 8% word error rate (WER) on average English speech. For natural conversational dictation, this is often acceptable. For technical terms, proper nouns, and non-English languages, accuracy may be lower.
Privacy
Section titled “Privacy”All processing happens on your device. Audio is never sent to Apple servers via Lagos. (Apple’s own Speech framework may use on-device or server-side processing depending on your macOS settings and internet connectivity.)
Language Support
Section titled “Language Support”Apple Speech supports the languages available in macOS’s Speech Recognition settings. You can select your language in Lagos Settings → Language.
Limitations
Section titled “Limitations”- ~8% word error rate (higher than WhisperKit or cloud engines)
- Accuracy varies with background noise
- Some languages may not be supported
When to Use Apple Speech
Section titled “When to Use Apple Speech”- You want instant dictation with zero setup
- Privacy matters and you want no downloads or API keys
- You need quick notes where exact accuracy isn’t critical
For higher accuracy, consider WhisperKit (Level 2).