Minimal pairs
Click a cell to compare clear and voiced.
or use the + on each cell
Minimal pairs in sentences
Two sentences side by side, differing by one sound in the same slot. Tap a card to hear the whole line, tap the highlighted part to hear just that.
Measured VOT
Shorter VOT = voiced; longer VOT = voiceless aspirated.
| Kana | Romaji | Type | Speaker | Duration | VOT | Voiced ratio |
|---|---|---|---|---|---|---|
| か | ka | clear | female | 190 ms | 32 ms | 63% |
| か | ka | clear | male | 160 ms | 132 ms | 12% |
| が | ga | voiced | female | 170 ms | 21 ms | 68% |
| が | ga | voiced | male | 180 ms | 37 ms | 61% |
| さ | sa | clear | female | 270 ms | 126 ms | 37% |
| さ | sa | clear | male | 180 ms | 117 ms | 22% |
| ざ | za | voiced | female | 230 ms | 1 ms | 67% |
| ざ | za | voiced | male | 230 ms | 17 ms | 76% |
| た | ta | clear | female | 130 ms | 16 ms | 68% |
| た | ta | clear | male | 150 ms | 42 ms | 23% |
| だ | da | voiced | female | 150 ms | 16 ms | 70% |
| だ | da | voiced | male | 180 ms | 17 ms | 86% |
| は | ha | clear | female | 220 ms | 66 ms | 57% |
| は | ha | clear | male | 180 ms | – | 0% |
| ば | ba | voiced | female | 160 ms | 2 ms | 81% |
| ば | ba | voiced | male | 190 ms | 37 ms | 62% |
| ぱ | pa | semi | female | 180 ms | 11 ms | 72% |
| ぱ | pa | semi | male | 140 ms | 52 ms | 21% |
Which pairs actually separate
Rule: a voiced VOT at least 10 ms shorter than its clear counterpart. Computed from the data — the female た/だ pair does not separate here, and the page says so.
| pair | female | male |
|---|---|---|
| か / が | separates | separates |
| さ / ざ | separates | separates |
| た / だ | no gap | separates |
| ば / ぱ | no gap | separates |
How to practise
- Hand on throat: か starts with airflow then voicing; が starts voicing almost at once.
- Mandarin ga is close to が, but が has a shorter closure.
- The は row is fricative; the real bilabial stops are ば / ぱ.
Where these numbers come from
VOT = speech-span start → first voiced frame; voiced ratio = voiced frames / frames in the span, both from parselmouth on normalised synthesis.
Synthesis VOT is an engine value: compare relatively; real speech varies with rate.