Urdu

Urdu ships five layouts. اردو is the phonetic key arrangement (Urdu letters sitting where the roman letter that sounds like them would be), and it is what you get when you add Urdu. Urdu phonetic is the new one (issue #496): a QWERTY grid that turns roman Urdu into the Urdu script as you type, so aap kaise ho commits آپ کیسے ہو. کلیدی تختہ (NLA) is the national standard keymap, اردو (four rows) spreads the letters over four rows, and اردو T9 is the keypad. This page is mostly about the phonetic one, because it is the one with moving parts.

Turning it on

WM Keyboard › Languages & layouts › اردو · Urdu › Layouts › Urdu phonetic

Add Urdu, open its screen, and tap the Urdu phonetic card under Layouts to turn it on. Tap the اردو card on the same screen to turn that one off if you don’t want both. Then do the one thing that makes the biggest difference: download the Urdu word list from the Dictionary row on that screen. Urdu ships no word list inside the app, and the next two sections explain why this layout cares about that more than most.

How it reads what you type

Every key is where it is on the English layout. What you type shows up in the Urdu script in the composing text straight away, with no separate conversion step, and the space commits it. The keys stay left-to-right while the text runs right to left, which is how every roman-key method for an RTL script works.

You type	You get	Why
din, kal, tum, mard	دن, کل, تم, مرد	a short vowel is a diacritic nobody types, so it writes nothing
ka, ki, ke, hai, to	کا, کی, کے, ہے, تو	a vowel that ends a word is long, because Urdu ends a word on a letter
ab, aap, ek, aur	اب, آپ, ایک, اور	a vowel that opens a word takes its carrier: ا, آ, ای, او
koi, hua, jaao	کوئی, ہوا, جاؤ	a vowel running into a vowel takes a hamza
yeh, woh, kya, liya	یہ, وہ, کیا, لیا	…but a ی or و that is a consonant carries nothing
dekh, likh, phir, ghar	دیکھ, لکھ, پھر, گھر	aspiration is do-chashmi he: ھ
khush, khabar	خوش, خبر	kh opening a word is خ, not کھ
sunna, pakka, iss	سنا, پکا, اس	a doubled consonant is written once: Urdu doubles with a shadda nobody types
accha	اچھا	…and a doubled c is the aspirate, which is the word people mean
Theek, kaRa, DaDa	ٹھیک, کڑا, ڈڈا	capitals spell the retroflexes exactly
Khabar, GhalaT, SubH, Zaroor	خبر, غلٹ, صبح, ضرور	and the Perso-Arabic letters: خ غ ح ص ض ط ظ ق
Kitab	کتب	a capital that spells nothing reads as its lowercase self

Urdu phonetic key map

Vowels. Urdu has no vowel signs. A vowel is a letter, a diacritic, or nothing at all, so the form depends on where in the word it falls.

Key	Opening a word	Inside a word	Closing a word
a	ا	nothing	ا
aa, A	آ	ا	ا
i	ا	nothing	ی
ee, ii, I	ای	ی	ی
u	ا	nothing	و
oo, uu, U	او	و	و
e, ay, ey	ای	ی	ے
ai	ای	ی	ے
o	او	و	و
au, ou	او	و	و
ua	او ا	وا	وا
ui	اوئی	وئی	وئی

Consonants.

b ب	bh بھ	p پ	ph پھ	f ف
t ت	th تھ	T ٹ	Th ٹھ
j ج	jh جھ	ch چ	chh, cch چھ
d د	dh دھ	D ڈ	Dh ڈھ
r ر	R ڑ	Rh ڑھ	l ل
s س	S ص	sh ش	z ز	Z ض
k ک	kh کھ / خ	q ق	g گ	gh گھ
Kh خ	Gh غ	h ہ	H ح	zh ژ
m م	n ن	N ں	v, w و	y ی
' ع	x کس	c ک

N is the nasalisation ں, which the rules never guess at (جان and ہیں are both “n” in roman), so type it when you want it, or let the word list settle it. A capital that isn’t in these tables reads as its lowercase letter.

Why Urdu is harder than Avro or Hindi phonetic

Urdu’s script does not write its short vowels. کتاب has no letter for the “i” of kitab and دن none for the “i” of din, so this transliterator writes fewer letters than it is given, the opposite of its Indic cousins, and the letters it does write are only the long vowels. Which leaves a pile of things roman spelling does not carry:

• whether a lone a is a long ا or nothing at all (kitab is کتاب, karna کرنا);
• whether a t is ت or ٹ, a d د or ڈ, an r ر or ڑ;
• which of the Arabic letters a loanword uses (khas is خاص, zaroori ضروری, galat غلط, subah صبح);
• whether a final a is ا or the ہ of a Perso-Arabic noun (bacha is بچہ, karna کرنا);
• whether a final n is ن or the nasalisation ں (jan is جان, hain ہیں).

The layout handles that in three layers, in this order.

1. A list of known spellings, about 580 of them. The commonest words outright (hai is ہے, hain ہیں, mein میں, kuch کچھ, bahut بہت, acha اچھا, pani پانی, theek ٹھیک), chat shorthand (k is کہ, h ہے, nhi نہیں, kr کر, bht بہت), and English loanwords written in the Urdu script (school is اسکول, doctor ڈاکٹر, mobile موبائل). A spelling on the list is settled there, and it shows in the composing text the moment you finish typing it.
2. The Urdu word list, if you’ve downloaded it. What you typed and every word on the list are both reduced to the word’s consonant skeleton. That is how the Arabic script spells in the first place, and why it fits. The skeleton forgets every vowel, forgets aspiration (خ and کھ are one key), folds ت ٹ ط together, and د ڈ, ر ڑ, س ص ث, ز ذ ض ظ, ک ق, ہ ح. Words with the same skeleton are candidates, and what you did type ranks them: a vowel at the start or end of a word must find one written, a vowel inside it need not, and a capital that asked for ٹ or ق and did not get it counts heavily against a candidate. Habits are forgiven (a closing ہ read as “a”, a ں left off, a long vowel typed with one letter), so hai finds ہیں as well as ہے, and bacha finds بچہ.
3. The rules alone, the table above. They write the literal reading: short vowels dropped, nothing retroflex, nothing Arabic.

With the word list installed, layer 3 rarely decides anything. Without it, layer 3 decides every word that isn’t on the list in layer 1, so expect kitab to read کتب and pani پنی until you download the list. That is why the strip also carries the other readings of what you typed: the same spelling with its last inner “a” long, with every one long, and with the first one long. Tap the one you meant.

The Urdu word list has no frequencies

Every word in the downloadable Urdu list ships with a count of 1, so there is no “commoner word” for the keyboard to prefer when two of them fit what you typed equally well. It falls back on the shorter word, which is right more often than not and sometimes isn’t. The spelling map in layer 1 is what covers the words you actually type most, and it is the thing to grow when something common comes out wrong.

Glide typing

Swiping over this layout draws a romanization, not an Urdu word, so the stroke is decoded against romanized spellings and the result is turned into the Urdu script. The known-spellings list is always there for that, and the Urdu (Romanized) word list (about 110,000 romanized spellings, in the language catalogue) widens it a lot if you download it. See Glide typing.

Editing

Backspace inside a word you’re still typing removes one roman letter per press, and the Urdu redraws. Over text that’s already committed it removes one code point, which on an Urdu word is one letter.

Options

WM Keyboard › Languages & layouts › اردو · Urdu

• Layouts: اردو, Urdu phonetic, کلیدی تختہ (NLA), اردو (four rows) and اردو T9, one card each, turned on and off independently.
• Key hints: WM Keyboard › Languages & layouts › Urdu › Key hints › Show the letter each key types. Off, Added, or Whole letter. It puts the Urdu letter a key is about to write in its corner. The same shared setting Avro and Hindi phonetic use, and it shows only while a transliterating layout is active.
• Spellings: WM Keyboard › Languages & layouts › Urdu › Spellings › Use known spellings, on by default. Switch it off for the letter-for-letter reading, which on Urdu means a lot more wrong guesses than on the other two.
• Dictionary words: WM Keyboard › Languages & layouts › Urdu › Spellings › Correct to dictionary words, on by default. Off, space commits exactly what the rules spell instead of a word from the list that matches its skeleton.
• The words before: WM Keyboard › Languages & layouts › Urdu › Spellings › Use the words before, on by default. One spelling can stand for several words, and the one or two words you typed just before pick between them, on the strip and for what space writes: “kam” is کام after “mera” and کم elsewhere. The evidence is your own typing and, where the language has one, the downloaded word-pair data. A known spelling only trades places with another known spelling of the same buffer, so the composing text is still what space writes. Turn it off if a word comes out differently after certain words than you expect.
• Also suggest from: with English picked, the phonetic layout’s strip also carries the word as you typed it in roman letters, and an English words group appears below it. See English on Avro and Hindi phonetic, which works the same way here.
• Numerals: ۰-۹ or 0-9.
• Dictionary: the downloadable Urdu list has about 155,000 words. You can import your own as well, and imported words are found through the same skeleton. The skeleton index over the downloaded list comes ready-made from the data repository (ur_phonetic_v1r1.wmpi.gz, about 1.5 MB, fetched right after the list) and is read in place from storage, so the keyboard no longer builds it in memory at every start. Only your imported words are still folded on the phone.

Details & edge cases

The اردو layout is also called “phonetic”, and is a different thing. Its keys are Urdu letters placed where the roman letter that sounds like them sits: a key arrangement. Urdu phonetic has roman letters on the keys and writes Urdu. Both are on Urdu’s screen; turn on whichever you meant.

kh is خ at the start of a word and کھ everywhere else. Both readings are everywhere in Urdu (خوش, خبر, خدا against کھانا, کھیل), but the aspirate is far commoner at the end of a word, where Urdu’s own verbs put it (دیکھ, لکھ, رکھ, آنکھ), and the Perso-Arabic letter is commoner at the start. Kh is خ wherever you type it. gh goes the other way and stays گھ, because گھر alone outweighs the غ words; Gh is غ.

The composing text can change its mind as you type. k is ک, ka is کا with the ا that closes a word, and the r of kar takes that ا straight back out again (کر). What you see at each point is what a space would commit at that point.

The phonetic machinery only loads if the layout is on. Urdu with just the اردو keyboard enabled builds none of it.

There are no selection-bar chips for Urdu yet. Bangla and Hindi have To Bangla / To Banglish and To Hindi / To Hinglish on the selection bar for converting text you already have. Urdu has the machinery behind them (both directions are implemented) but no chip of its own; ask on the issue if you want one.

It’s new, and both tables are a draft. The rules, the spelling map and the English loanword list were written without a native speaker on hand, and measured against the word list rather than reviewed: the rules alone get about 62% of a 176-word test set exactly right, and the skeleton index puts the right word first for 44% of them and in the first three for 73%. If a common word comes out wrong, say so on the issue and name the word and the spelling you typed.