Apple Speech

When you choose the Apple Speech engine in Action Phrase, the app uses Apple’s built-in on-device speech recognition to convert your voice into text. This engine is optimized for speed, privacy, and tight integration with iOS and iPadOS.

Below are the available configuration options.

Language

The Language setting determines which speech recognition locale Apple’s engine uses when listening.

What It Does

• Selects the speech recognition model for a specific language and region (for example, English (US), English (UK), Mandarin (Taiwan), etc.).
• Affects pronunciation detection, word matching, and overall recognition accuracy.
• Impacts how phrases are interpreted, especially with accents or regional vocabulary.

When to Change It

• You regularly speak in a language other than your device’s system language.
• You present or perform in multilingual environments.
• You want higher accuracy for a specific accent or regional dialect.

Best Practices

• Match the language setting to the language you are actively speaking.
• If you frequently switch languages during events, consider organizing phrases into groups aligned with the active language.
• Keep your phrase text written naturally in the same language you’re speaking.

Debounce Partial Transcript

The Debounce Partial Transcript option controls how Action Phrase evaluates live, in-progress speech recognition results from Apple’s Speech engine.

Understanding Partial vs. Final Results

Apple’s speech engine produces:

• Partial transcripts — live, continuously updated text while you are still speaking.
• Final transcript — the completed, stabilized recognition result after a pause or when speech has settled.

In fast-moving environments, you may not always receive a final transcript immediately — especially if you’re speaking continuously.

What Debouncing Actually Does

When Debounce Partial Transcript is enabled, Action Phrase:

• Waits briefly before acting on partial transcripts.
• Gives the speech engine time to stabilize the recognition.
• Falls back to evaluating the most recent partial transcript if a final transcript does not arrive in time.

This means:

• The app prefers final results for accuracy.
• But it will not wait indefinitely.
• If no final transcript is delivered quickly enough, the latest partial transcript is evaluated so your phrase can still run.

This provides a balance between accuracy and responsiveness.