Speaking is sometimes simply faster

There are people who type slowly, people who think of something worth keeping while walking, and people who look at a long reply they have to write and cannot face the keyboard. Voice input exists for all three. This keyboard comes from a company that has spent years on speech technology, and it puts that work directly under the thumb: hold the voice key, talk, and the characters appear on screen as they are spoken. The official page advertises a rate of 400 characters in a minute, which sounds like marketing until somebody who types slowly tries it once.

Recognition is the part that had to be right

Speech input handles Chinese and English mixed together, so an English word dropped into a Chinese sentence is transcribed rather than mangled. Dialect speakers are looked after as well: several common varieties are supported directly, which matters for anyone whose typing is slower than their speech because the keyboard was designed around a different accent. Recognition appears while the sentence is still being spoken, punctuation is placed automatically, and the cost of correcting afterwards stays low. In a quiet room the result is genuinely good; in a noisy one there is noise processing underneath to keep it usable.

Where voice stops being the better choice is worth knowing too. In a room with a television on, or on a train, recognition has to work harder and the correction afterwards costs more than typing would have. For a name, a code or a single unusual word, the thumb is still faster. The sensible habit is not to make voice the only way in but to reach for it when the sentence is long and the thought is already formed, which is exactly the case the keyboard was built around.

What the 2026 releases added

The 2026 versions connect the keyboard to the company own large model, and two changes follow. The first is an assistant that works inside the keyboard: polishing a piece of text, drafting a reply in the right tone, or answering a question without leaving the app being typed into. The second is a shelf of custom skills, where the functions a person uses most, along with assistants and stock phrases, can be arranged freely and switched as a set when the situation changes, so that the things done ten times a day are one tap away instead of three.

One more thing about the rate on the poster. Four hundred characters a minute only describes somebody speaking fluently into a quiet phone, and it is not a promise about a finished document, since a dictation still has to be read once afterwards. What changes is not the peak speed but the fact that a first draft, usually the hardest part of writing anything, stops being the reason a piece never gets started.

The ordinary input is not an afterthought

Voice is the headline, but the rest of the keyboard is solid. The pinyin engine covers the full spelling scheme, the double-stroke scheme and fuzzy sounds, and a cloud dictionary keeps the vocabulary of the moment up to date. Handwriting accepts characters written over one another and joined together, which is the fastest way out when a rare character will not come to mind. On top of that sit the cosmetics and the small tools: a large collection of skins, fonts and key sounds, clipboard management, and stored phrases for the messages sent every day.

Several devices, one dictionary

There are versions for computers and for phones, and the account and the dictionary stay the same across them, so a phrase learned at the desk is available on the phone an hour later. As of 2026-09 the Windows release is numbered 3.0.1750, the mobile line has reached its fifteenth series, and the assistant is being rolled out across the platforms gradually. The software is free for an individual, and the package is distributed from the official site.

Who benefits most

Three groups notice the difference quickly. People who write for a living can dictate a rough draft of a long piece and come back to polish it, which changes how much of a day goes into production. Heavy users of chat get replies improved by the assistant, in a tone that suits the conversation. And for dialect speakers and older relatives, the barrier disappears: anybody who can talk can type, and nobody has to learn a keyboard layout first. For anyone who has never made voice the main way of entering text, the official site is the place to look, and one attempt usually settles the question.