How Words and Characters Are Counted in Hindi Text

Type नमस्ते into a form that allows “10 characters” and you have used 6 of them, not 3, even though you see three letters on the screen. Paste a Hindi paragraph into a tool built for English and the word count may be wrong too. Neither the form nor the tool is broken; they are counting different things. This guide explains the three ways a computer can count Hindi characters, which one each situation uses, and how word and sentence counts work for Devanagari. Every number comes from the same function that runs our word and character counter, which was built for Hindi and other Indic scripts.

Why Hindi is different from English

An English letter is one unit on every level: one key, one code, one visible shape. Devanagari is an abugida: a consonant carries an inherent vowel, other vowels are written as matras attached above, below, before or after it, a virama (halant, ्) removes the vowel and joins consonants into conjuncts (संयुक्ताक्षर), and signs like anusvara (ं), chandrabindu (ँ), visarga (ः) and nukta (़) modify the letter further. A single visible unit such as स्ते is built from several codes: स + ् + त + े.

So the question “how many characters?” has three honest answers.

The three counts

1. Grapheme clusters: what a reader sees. One cluster is one visible letter-unit: a consonant with all its matras and marks, or a complete conjunct. This is what our counter calls characters. नमस्ते = न, म, स्ते = 3.
2. Code points: what Unicode stores. Each consonant, matra, virama and sign is its own code point. नमस्ते = न म स ् त े = 6.
3. UTF-16 code units: what most forms and databases count. For Devanagari every code point is one code unit, so this equals the code-point count: नमस्ते = 6. Our counter shows this as code units, because it is the number maxlength limits, SMS counters and most “characters remaining” boxes use.

Read down the table. क्षत्रिय is three visible units (क्ष, त्रि, य) but eight code units. स्वतंत्रता is four units but ten. For ToolJagat the two counts agree, which is why English-only tools never had to make the distinction.

Matras, conjuncts and signs, one by one

• Matras (ा ि ी ु ू े ै ो ौ) never count as separate characters to the eye; they attach to the consonant before them. हिंदी = हिं + दी = 2 characters, 5 code units.
• Anusvara and chandrabindu (ं ँ) attach to the syllable they sit on. हिंदी (with anusvara) and हिन्दी (with a half न conjunct) are both 2 characters, but 5 and 6 code units respectively. Both spellings are correct Hindi; they are stored differently.
• Virama conjuncts: the virama glues consonants into one cluster. ज्ञान = ज्ञा + न = 2 characters; श्री = 1 character and 4 code units.
• Nukta (़, the dot under ड़ ढ़ क़ ख़ ग़ ज़ फ़) is a combining sign and stays inside the cluster: पढ़ाई = 3 characters. Some keyboards produce the precomposed letter (one code point) and others the base letter plus nukta (two); the visible count is the same, the code-unit count differs by one per nukta.
• Danda (।) and double danda (॥) are punctuation: they count as a character and end a sentence, like a full stop.

A note on where the counts come from: our counter uses the browser’s own Intl.Segmenter, which follows Unicode’s text-segmentation rules (UAX #29). Recent Unicode versions treat a virama-joined conjunct as a single cluster; older browsers and devices may split a conjunct into two clusters, so the “characters” figure can differ slightly between devices for conjunct-heavy text. The code-unit count never varies.

How words are counted

A word is a run of letters and digits between boundaries (spaces, punctuation, line breaks), found with the same Unicode rules. Some consequences:

• The sentence भारत की राजधानी नई दिल्ली है। has 6 words, 19 characters including spaces and 14 without.
• Punctuation is not a word: the danda is not counted, and neither are commas or question marks.
• A hyphenated or slash-joined pair may count as two words depending on the rules; a number like 1,400 counts as one. The line दिल्ली से मुंबई 1,400 km है is 6 words.
• Numbers written in digits count as words; numbers written in words count as however many words they contain.

Sentences, paragraphs and reading time

Sentences end at a full stop, question mark, exclamation mark, danda or double danda followed by a space or line end: मैं रोज़ सुबह दौड़ता हूँ। क्या आप भी? is 2 sentences, 8 words. Paragraphs are separated by a blank line. Reading time divides the word count by 200 words a minute and speaking time by 130, the usual planning figures; Hindi read aloud at a natural pace is in that range for most speakers, but adjust for your audience.

When each count matters

• Form fields, usernames, SMS, passwords, tweet-length limits: code units. A “100 character” box holds fewer visible Hindi letters than English ones, often 60 to 70 depending on how many matras and conjuncts the text has.
• Typing speed tests and transcription: usually visible characters, sometimes key presses (which are closer to code units). Check which the test uses before practising.
• Essay limits in competitive exams: words. Count with the word rule above and remember that the examiner counts visible words, not characters.
• Translation and content pricing: words for most agencies, sometimes characters for short strings. Agree the unit in writing; Hindi text runs 10 to 20 percent longer in words than its English source for most prose.
• Database columns and printing: code units for storage, grapheme clusters for layout width; a VARCHAR(50) column holds about 25 to 30 Hindi letters.

Common mistakes

• Trusting an English word counter for Hindi. Tools that split only on spaces miscount around dandas and punctuation, and those that count bytes report numbers three times too large (each Devanagari code point is three bytes in UTF-8).
• Assuming one visible letter is one character in a form. It is usually two or three code units; plan for it when a limit is tight.
• Comparing counts from two spellings. हिंदी and हिन्दी look alike and read alike but store differently; a search or a duplicate check that compares code units treats them as different words.
• Expecting identical character counts on every device. For conjunct-heavy text the grapheme count can vary by browser; the code-unit count is stable, so quote that for limits.

Frequently asked questions

How many characters is नमस्ते? 3 visible characters (grapheme clusters), 6 code units. A form with a character limit counts the second figure.

Do matras count as characters? Not to a reader, and not in our “characters” figure: a matra joins the consonant before it. They do count in code units, which is what form limits use.

Why does my form say the limit is exceeded when I can see fewer letters? Because the form counts code units. Each matra, virama and sign is one unit even though it is not a separate visible letter.

Does a danda count as a word? No. It is punctuation: counted as a character, used to end a sentence, never a word.

Why does the character count differ between my phone and my laptop? Different Unicode versions on the two devices may split or join conjuncts differently. The code-unit count is identical everywhere.