The 46 basic kana
Read across a row to keep the consonant; read down a column to keep the vowel. Blank cells are sounds Japanese does not have.
Four that break the romanisation
Romanisation is a transliteration, not English spelling.
not 'si'; closer to Mandarin xi, but shorter
closer to Mandarin q, without the retroflex
like Mandarin c plus a very light u
bilabial [ɸ], not the English labiodental f
Why every character is one mora
Japanese counts time in morae, not syllables. Each kana is one mora, and the relative length of morae stays constant at any speaking rate — that is where the rhythm comes from.
A contracted sound (きゃ, しゃ, ちゃ…) is two characters but one mora; っ, ー and ん each take a full mora.
How to use this page
- ① Down a column: same vowel, different consonants.
- ② Across a row: same consonant, five vowels — keep the vowel quality steady.
- ③ Click a cell: each clip is synthesised in isolation, without sentence intonation.
Where the numbers come from
The chart, romanisation and mora counts all come from one table in 50_ja_mora.py; audio is synthesised per character with ja-JP-NanamiNeural. Nothing is hard-coded.