· 4 min read
WordPace: measuring speaking pace on device
Speaking too quickly is one of the most common items in presentation feedback, and it usually arrives after the talk; while actually speaking, especially under nerves, people misjudge their own rate. WordPace moves that feedback into the talk itself. It measures speaking rate on iPhone and Mac against a configured limit in words per minute and displays a warning the moment the rate exceeds it.
Display
The current rate is shown in large type. When it exceeds the configured limit, the display fills with a warning reading SLOW DOWN, which clears again on its own once the rate is back below the limit. There is no score, no evaluation after the session and no completion metric. During a talk the device is only in peripheral vision, where a colour change on a numeric value is not reliably perceived, whereas a change of the entire display area is; the visual design follows from that constraint rather than from a stylistic preference.
Measuring over speaking time
Words per minute computed over wall-clock time treats pauses as reduced pace. A deliberate pause after a key sentence then lowers the measured rate, and once speaking resumes, the average takes a correspondingly long time to reflect the actual rate, so the warning arrives late. WordPace therefore measures over speaking time: a rolling window of roughly twenty seconds of detected speech, preceded by a short warm-up so the first value is not derived from two words. The warning has hysteresis — it clears only once the rate is clearly below the limit, not at the boundary itself — which prevents it from toggling repeatedly around the threshold.
Languages
A comfortable speaking rate is a property of the language rather than of the speaker, so a single limit does not transfer between languages: German, English and Japanese sit at different values. WordPace supports 45 presentation languages, stores a separate limit for each and supplies a default per language. The interface itself is localised into eleven languages. The speech model for a presentation language is downloaded once on first use; the measurement then runs offline.
Audio handling
The microphone signal is processed on the device. Once the rate has been computed, the audio is discarded: no recording is stored, no transcript is kept, and the application has no network access of its own. What persists is the configuration — limits, presentation language, display options. This follows the same reasoning as in Clessira: the measurement is the result that is needed, and the underlying data has no purpose once it has been computed.
Requirements
WordPace requires macOS 26 or iOS 26 and is a single purchase covering both platforms, without a subscription or an account. On macOS the warning is drawn over ordinary windows but not over another application's full-screen presentation, which is relevant when the presentation runs on the same machine. On iOS the warning stays within the app, which matches the usual setup with the device placed on the lectern. A demonstration of the readout is available at wordpace.de.