Arabic Tashkeel Explained: Adding Diacritics by Hand and Automatically

Tashkeel (تشكيل) is the set of small marks written above and below Arabic letters to show the short vowels and other pronunciation details: fatha, damma, kasra, sukun, shadda and tanween. Most Arabic is written without them, because native readers infer the vowels from context. You need them when a word could be read two ways, when the text is for learners, and when a voice actor or a text-to-speech engine has to read it aloud.

This guide covers what each mark does, when to vocalize a text fully or partially, how to type the marks, and where automatic diacritization tools still go wrong.

The tashkeel marks

A shadda can carry one vowel on the same letter, as in مُدَرِّس (mudarris, teacher). The vowel on the last letter of a word depends on its grammatical role in the sentence, which is called i'rab (case endings). That's the hardest part of tashkeel for people and for software alike.

When does Arabic text need tashkeel?

Ask one question: could this word be read two different ways here? If so, mark it.

• Words spelled the same: عَلَم ('alam, flag) and عِلْم ('ilm, knowledge); قَدَّمَ (qaddama, presented) and قَدِمَ (qadima, arrived). One mark usually settles it.
• Names of people and places: عُنَيْزَة (Unaizah), الْقُرَيَّات (Al-Qurayyat), يَنْبُع (Yanbu). Readers from elsewhere, and software, often get these wrong.
• Teaching material and children's books: vocalize fully, because the reader is learning the pronunciation itself.
• Poetry, religious text and quotations: a wrong vowel can break the meter or change the meaning.
• Scripts that will be read aloud: whether by a voice actor or by text to speech, tashkeel settles the pronunciation instead of leaving it to a guess.

Full vocalization of every letter makes on-screen reading slower and multiplies typos. For articles, ads and customer messages, partial tashkeel, marking only the ambiguous spots, is the norm.

How to type tashkeel

On the standard Arabic (101) keyboard in Windows, each mark is Shift plus a letter key. Type the letter first, then the mark:

These come from the keyboard layout, so they work the same in Word, Google Docs and the browser. The Mac Arabic layout puts them elsewhere: open the Keyboard Viewer and hold Shift to see them.

Automatic tashkeel: how it works and where it fails

Automatic diacritizers read the whole sentence and predict each letter's vowel from context. Known options include Mishkal (open source), Farasa from the Qatar Computing Research Institute, and general-purpose AI models, which will add tashkeel if you ask. You can also try our free Arabic tashkeel tool, which runs in your browser.

How accurate they are depends on where in the word the mark goes:

• Inside words: most accurate. A good tool rarely misses the vowels of مَدْرَسَة (madrasa) or مُعَلِّم (mu'allim).
• Word endings (case endings): most errors land here, especially in long sentences or when the subject comes late.
• Proper nouns and loanwords: a tool can't know how people in Al-Ahsa say their city's name, or how you want your product name pronounced.
• Dialect: case endings don't apply to spoken dialects, and most tools are trained on Modern Standard Arabic. وش رايك (wish rayik, "what do you think" in Najdi) shouldn't be vocalized like MSA.

General-purpose AI models carry one more risk: they sometimes change a word, or "fix" what they think is a typo, while adding the marks. Compare the letters with your original before you use the result.

Tashkeel and Arabic text to speech

A text-to-speech engine can't see vowels you didn't write, so it has to predict them before it speaks. That's why Arabic TTS quality varies more than in most languages: many errors you hear aren't in the voice at all, but in a wrong vowel guessed a step earlier.

Here is a fully vocalized Modern Standard Arabic sentence with a place name and a date. In English: "Diriyah is considered the cradle of the first Saudi state, and the At-Turaif district was inscribed on the World Heritage List in 2010."

Dialect is vocalized differently. In Najdi, word endings are usually silent (sukun), and some words have no MSA equivalent (more on how the Saudi dialects differ in our guide to Saudi Arabic text to speech). This one says: "Hey! Shall we start? Type any text and leave the voice to us."

Samar adds tashkeel automatically before reading, so you type unvocalized Arabic as usual. If a word comes out with a pronunciation you don't want, you change its vowel with a click instead of rewriting the sentence. A pronunciation dictionary keeps your brand and product names right once you've set them.

Common tashkeel mistakes

1. Vocalizing every letter of text meant for normal reading: it weighs the text down and doubles the chance of errors. Mark only what's ambiguous.
2. Copying vocalized text out of a PDF: many PDFs detach the marks from their letters or reverse their order on copy. Check the text after pasting.
3. Mixing up ة and ه: not tashkeel strictly, but it changes pronunciation. مدرسه with ه can be read mudarrisuhu ("his teacher") rather than madrasa (school).
4. Leaving place names unmarked in a voice script: if the text will be spoken, mark every proper noun, even well-known ones.
5. Vocalizing dialect by MSA rules: كيفك ("how are you") in dialogue doesn't become kayfuka. Mark dialect as it's spoken.

The short version

Tashkeel exists to remove ambiguity, not to decorate text. Mark what might be misread, mark everything in teaching material, and check any machine output at names, numbers and word endings. If your text is going to become audio, getting the vowels right is half the quality of the result.

If your text is a video voiceover, the full workflow is in how to make an Arabic AI voice over.

Try a line of your own script in Samar's free demo and hear how it's vocalized and read before you publish.

Frequently asked questions

What does tashkeel mean in Arabic?

Tashkeel (تشكيل) means adding the marks that show short vowels and other pronunciation details: fatha, damma, kasra, sukun, shadda and tanween. Most everyday Arabic is written without them.

Can tashkeel be added automatically?

Yes. Tools such as Mishkal and Farasa, and general AI models, can vocalize text. They are accurate inside words but make mistakes on case endings, proper nouns and dialect, so check those by hand.

How do I type Arabic diacritics on a keyboard?

On the Windows Arabic keyboard, hold Shift: Q for fatha, E for damma, A for kasra, X for sukun and the backtick key for shadda. Type the letter first, then the mark.

Does Arabic text to speech need vocalized text?

Not necessarily: Arabic-first engines add tashkeel before speaking. Marking names and ambiguous words yourself guarantees the pronunciation you want.