Speech Rate Meter 2
Speech Rate Meter 2 is a cross-platform desktop and Android application that measures how fast you are speaking. A session stays open while you talk. Each pause-bounded phrase is measured from vowel-like intensity peaks. The microphone audio stays until you leave the app, unless Keep recording files is on.
It does not recognize words and does not use a speech-recognition model. Pace, articulation, pauses, and speech time are estimated on the device from the intensity contour of the recording.
This repository is a rebuild of the earlier Speech Rate Meter (Qt 5, separate Inton Core library). Version 2 keeps the pace measures, adds a live analysis window, and keeps only the numeric result. The filler score from an earlier build of version 2 has been removed.
Designed for speakers, announcers, call-center operators, language learners, and clinicians who want an objective, repeatable reading of speaking tempo — for rehearsal, self-control, or observation of oral speech.
Package id: by.intoncore.SpeechRateMeter2. Version name 1.0.1, version code 19.
What It Does
Core Concept
- Press Start. The microphone stays open for the whole session.
- With Detect speech automatically on, a pause ends the current phrase, the phrase is measured, and the next one starts when you speak again. With it off, measuring starts when you press Start, a pause does not end the phrase, and the take is split into 15-second parts.
- While you speak, speech rate, articulation, and pauses follow the analysis window — the most recent kept speech (10 seconds by default) — so a change of pace shows within a few seconds. Direct speech rate, direct articulation, and the pause length use that same window. They are not a running total from Start. Speech is the total time measured since Start.
- Press Stop. Home and History then show the selected summary of speech rate, articulation, and pauses for every phrase kept from Start to Stop. Speech is the total time. The screen names that summary. A phrase shorter than one second is dropped.
- The scratch WAV stays after that phrase’s numbers are stored, so Listen can play it after Stop. With Keep recording files on, Stop joins those phrase files into one WAV. Off, the phrase files are deleted when you leave the app. There is no waveform.
All processing is local. The Android package requests only microphone access (RECORD_AUDIO).
The measures
| On screen | Meaning | Unit |
|---|
| Speech rate | Overall pace. The gauge names it Slow, Average, or Fast | words per minute |
| Articulation | Pace while speaking, with the gaps left out. Never below the speech rate | words per minute |
| Pauses | Length of the longer gaps. Not the total time spent silent | seconds |
| Speech | Total time that counted as speech. A saved result shows the card when Show Speech tile is on. While recording the red badge shows elapsed time. Silence before speech is detected is left out. A pause counts only up to Pause (s). From one minute the card reads 1m 30s | seconds |
Speech rate is vowels per minute scaled into words per minute (K1, default 0.41). With Speech rate method set to Direct (the default), that time runs from the first non-silent frame through the last, so silence before the first sound and after the last is left out. Legacy uses the whole buffer. Direct articulation removes measured pauses from that same span. With no such pauses the two direct rates are equal. Legacy articulation keeps the older estimate from the typical gap between vowel nuclei. The full formulas are in docs/TECHNICAL_DESCRIPTION.md.
User Experience Flow
1. Home
Home is the only screen with the record button.
以上為 README 的開頭部分。其餘內容請閱讀原文。 檢視完整 README ↗
