9-Key, Handwriting and Full Pinyin in One Chinese Keyboard | Loopal

The short answer

Three layouts behind one keyboard. Loopal's Chinese extension lets you pick full Pinyin (26 letters, both hands), 9-key Pinyin (a three-by-three digit grid you can work with one thumb), or handwriting (draw the character) — and the choice is a layout setting inside the keyboard, not a different keyboard and not a different app. Pinyin candidates and handwriting recognition both run locally on your device with no network access. Installing it does mean adding two keyboards to iOS and deciding whether to grant Full Access, and Loopal has not shipped on the App Store yet.

Why one keyboard cannot cover all three

The problem behind the search is physical, not linguistic. You type Chinese in at least three different postures, and each one makes a different layout correct:

• A desk, or a tablet held in two hands — full Pinyin, and nothing else comes close.
• A train, a queue, one hand on the phone — the 9-key grid, because your thumb cannot cover 26 keys.
• A name, a place, an obscure surname — handwriting, because the character either has no common pinyin or has several, and guessing costs more than drawing it.

No layout is fastest at everything, so the useful thing is not a better single layout. It is a cheap switch between three. That is the whole difference between a keyboard that offers three layouts and a keyboard that offers one: with one layout, a keyboard extension forces you to enable a second keyboard and hit the globe key mid-sentence, so the switching cost is the thing you actually pay.

Full Pinyin: the mode you will use most

The letter plane is three rows of 26 letters with a function row of 123 | punctuation | space | 中/英 | return. Two details that are easy to get wrong and that tell you whether a keyboard was actually designed or merely configured:

• Gesture typing resolves only here. A swipe on the letter plane yields a string of letters, and the engine needs a syllable string to work with — so that is the only plane where it can. The symbol plane and the 9-key grid have no letters to swipe across at all, which means gesture input is a full-Pinyin affordance rather than a keyboard-wide one.
• Punctuation is tap-and-swipe, not tap-and-wait. Tap inserts the primary mark and swipe up inserts the other. In Chinese the comma and the period are near-equal in frequency, and forcing one of them to wait out a long press costs a delay at the end of every sentence. The long-press variants appear in a bubble you can slide away rather than a swap in place, because a bubble can be dismissed and a replaced character cannot be un-replaced.

The 中/英 key on that same row is a language switch, not a layout switch. It is orthogonal to the three layouts below it, which is why the two decisions never fight each other.

9-key: what the grid saves and what it charges

The grid is 1 | 2 ABC | 3 DEF, 4 GHI | 5 JKL | 6 MNO, 7 PQRS | 8 TUV | 9 WXYZ. Two details worth knowing before you switch to it:

• 0 is the syllable separator — the key labelled 0 分词, which is how you break one ambiguous run of digits into two syllables instead of letting the engine guess.
• The function row keeps its own backspace. On the letter plane the backspace rides at the end of the bottom letter row; on the 9-key grid there is no bottom letter row to carry it, so the function row grows by one key: 123 | punctuation | space | 中/英 | backspace | return.

What it saves is reach: nine keys instead of twenty-six, arranged three by three for a thumb. What it charges is keystrokes and candidates — more taps per syllable, and more candidate selection than full Pinyin produces for the same character. There are also no letters on this plane to swipe across, so the gesture typing described above simply does not exist here.

Handwriting: the escape hatch, and what it costs

Handwriting is a canvas you draw into, with an eraser, a recognize action and a horizontal candidate strip above it, plus a control that takes you straight back to the Pinyin keyboard. The recognition is a deliberate act: you draw the character, you press recognize, you pick from the candidates. Recognition itself runs through the system text-recognition stack — at accurate level rather than fast — entirely on the device, which is why the candidates appear without any network round trip.

The honest limit is that this is not a fast mode for prose. It is for the character that has no common pinyin: a surname, a street name, a technical term. It is deliberately slower, and treating it as a general-purpose input method is how it disappoints people.

What stays the same when you switch

• One panel owns the layout. Changing your typing method is an act you want at the moment you change it, not a preference buried among toggles, and having two controls that both claim to be the current value is a bug. So the layout lives in exactly one place, and every mode is one entry in the same three-card picker.
• The switch is internal. Switching layouts keeps the candidate bar, the 中/英 toggle and the keyboard itself. You are not leaving the conversation you are typing in, and you are not hunting for another keyboard in Settings.
• The dictionary ships once. The Chinese extension carries the compiled input data — about 52 MB — and the English keyboard reads that same copy out of the same app container instead of carrying its own. Shipping it twice would have added 52 MB of duplicate bytes to the install for nothing.
• No layout is a paid feature. Pinyin, the 9-key grid and handwriting all run on device, and none of the three is gated. What the Pro plan changes is the assistant's credit allowance, and early access to new skills and platform models.

What it costs to install

• Two rows in your keyboard list. Loopal 中文 and Loopal English are separate keyboard extensions, because iOS lets one extension declare only one keyboard service. Their display names are pinned rather than following your device language, precisely so you can tell the two apart in Settings.
• A Full Access decision. Full Access is what lets the keyboard reach the network and stay in sync with the main app, so it is what the in-keyboard assistant needs. Ordinary typing does not need it — and, as above, does not use it either.
• iOS 17 or later.
• Nothing yet, on the App Store. Loopal is still in development with no App Store ID, so the download link on its product page is a placeholder.
• Against what you already have. iOS ships its own Pinyin, 9-key Pinyin and handwriting keyboards for free. None of these three layouts is something you cannot already do; the question is whether one keyboard holding all three, plus something else, beats the built-ins plus a globe-key press.

What Loopal does differently

On the input side Loopal does not claim a better layout. It claims three of them behind one keyboard, one place to change them, and recognition that never leaves the device. What surrounds that is the part no built-in keyboard has: hold the space bar and the assistant is there, inside whatever app you were typing in, with the answer dropping into the field you were already in — and a backend you pick rather than one chosen for you (platform credits, your own Anthropic / OpenAI / Doubao / Gemini key, or an Ollama server on your own LAN). That is the reason to consider it, and the three input modes are what has to be solid for the whole thing to be worth anything.

Worth stating plainly: the assistant needs Full Access; the platform-credit path needs an account while your own key or your own Ollama does not; and none of it has shipped yet.

The decision you can make now

If you type Chinese two-handed at a desk and that is all, the built-in Pinyin keyboard already covers it and a third-party keyboard is a step backwards in simplicity. If you also type one-handed on a commute, or keep hitting a character with no common pinyin, three layouts behind one keyboard removes the globe-key press — and Loopal is worth following on agentsmesh.app/products/quipkey for the part the built-ins cannot do: a conversational assistant that lives in the same keyboard and talks to a backend you choose. Budget for two keyboard rows, a Full Access decision, iOS 17, and a wait. If the assistant is not the part you want, the input modes alone are not a reason to install anything.