What the dictation speed test measures
The timer measures how quickly words reach the field, not only how fast you speak. That includes recognition delay, your start and stop habit, and any pause before finishing.
The word score compares normalized word tokens with the fixed reference passage. Insertions, deletions, and substitutions each count as one edit. Capitalization and punctuation are excluded so the score stays focused on recognized words.
How to compare tools fairly
Keep the Mac, microphone, room, passage, language, and distance from the microphone the same. Run several attempts per tool and compare the middle result rather than the single fastest score.
A short passage cannot prove universal accuracy. Names, technical vocabulary, accents, long-form fatigue, field insertion, and network conditions can change real-world performance.
- Use the same reference passage and hardware.
- Do not correct one tool and leave another uncorrected.
- Record at least three attempts under the same conditions.
- Judge the workflow you actually use, not only one score.
Why this page does not open your microphone
Browser speech APIs can introduce another speech engine and another processing boundary. That would test the browser rather than the Mac dictation product you meant to evaluate.
IraVoice therefore leaves microphone capture to your chosen tool. The test receives only the temporary text you place in the field, calculates the result locally in the page, and discards it when you reload or close the tab.