Japanese Voice Typing

Typing Japanese normally has two stages: enter the sounds as kana, then choose which kanji those sounds should become from a list of candidates. The second stage is the work.

Dictation collapses it. The recogniser decides from context which kanji you meant and writes them directly, so you never see a candidate list. For fluent speakers this is dramatically faster than typing. The cost is that the choice is no longer yours — and when the choice is wrong, you are correcting text rather than picking a different candidate.

Japanese has a great many words that sound identical. はし is 橋 (bridge), 箸 (chopsticks) or 端 (edge). こうしょう has more readings than most people can list.

Context resolves these correctly most of the time. When it does not, the result is a perfectly grammatical sentence about the wrong thing — which is the hardest kind of error to catch, because nothing looks broken. Proofread dictated Japanese for meaning rather than for form; scanning for typos will not find anything, because there are none.

Japanese is written without spaces between words. The reader’s cues are the alternation between kanji and kana, and the punctuation.

Sentence-ending 。 is usually inserted for you. The comma 、 is much less reliable, and a long unpunctuated Japanese sentence is harder to read back than the equivalent in a spaced language. Dictating a sentence at a time and adding 、 as you go is worth the interruption.

The microphone is your browser’s. On Chrome and Edge the audio is sent to the browser vendor’s speech service and text comes back; WorldTyping is never in that path and never receives audio. The text stays in your browser, saved locally so a reload does not lose your draft.

That is also why dictation needs an internet connection, and why a browser without Japanese speech support will not respond to the button.