Turkish Text-to-Speech: İ and ı, Suffixes and Stress Explained

Turkish is in one way a friendly language for text-to-speech: its spelling is close to phonetic, so each letter almost always stands for the same sound. The problems lie elsewhere: the dotted and dotless i (i/İ and ı/I) get mangled by software and by people typing without Turkish letters, suffixes attached to numbers and names have to match how those are pronounced, and word stress falls in places the spelling doesn't show. Handle those three things and Turkish synthetic speech sounds very natural.

Is Turkish easy or hard for text-to-speech?

Both. The Latin-based alphabet adopted in 1928 has 29 letters, and the match between letters and sounds is far more regular than in English or French. A voice rarely has to guess how a correctly spelled Turkish word sounds.

The difficulties come from four places:

1. Text that isn't correctly spelled, especially missing ı, ş, ç, ğ, ö and ü.
2. Suffixes on numbers, abbreviations and foreign names, which follow pronunciation rather than spelling.
3. Stress, which is usually on the last syllable but has important exceptions.
4. Long, suffix-heavy words and sentences that need sensible phrasing.

Why do the dotted and dotless i matter so much?

Turkish has two different i vowels, each with its own capital letter:

Lowercase	Uppercase	Sound
i	İ	like the "ee" in "see", but shorter
ı	I	an unrounded vowel made further back, roughly the vowel at the end of "cousin"

They are separate letters with separate meanings. Kır is countryside; kir is dirt. Işık (light) and ılık (lukewarm) have no dotted i at all. The problem is that text often reaches a voice with the wrong one, in two common ways.

Text typed without Turkish letters

On older systems, foreign keyboards and in quick messages, people often write "sikinti" for sıkıntı (trouble), "gunes" for güneş (sun) and "cok" for çok (very). A human reader fills in the right letters automatically. A voice reads what is there. Because the Turkish letter c sounds like an English "j", "cok" comes out as "jok", which is not the word you meant. If your text comes from a database, a spreadsheet export or a translation tool that strips special characters, restore the Turkish letters first.

Capital letters converted with the wrong rules

In Turkish, i becomes İ in capitals and I becomes ı in lowercase. Software that uses English case rules gets this wrong: the heading KIRMIZI (red) lowercased the English way becomes "kirmizi" instead of kırmızı, and istanbul in English capitals becomes ISTANBUL instead of İSTANBUL. If you paste headings written in capitals or text that has passed through a case conversion, check every i. Better still, write scripts in normal sentence case.

How do vowel harmony and suffixes affect Turkish speech synthesis?

Turkish builds words by adding suffixes to a root, and the suffix vowels change to match the vowels of the word: this is vowel harmony.

• Two-way harmony: the plural is -lar after back vowels and -ler after front vowels: kitaplar (books), evler (houses).
• Four-way harmony: some suffixes choose between ı, i, u and ü: kızı (the girl), evi (the house), okulu (the school), gözü (the eye).

In correctly written text the harmonized vowel is already on the page, so the voice doesn't have to work it out. It matters in three places where the spelling and the sound part ways.

Suffixes on numbers

Turkish attaches suffixes to digits with an apostrophe, and the suffix follows how the number is said. 2026'da (in 2026) is read iki bin yirmi altıda. 6'da is altıda, but 4'te is dörtte, because dört ends in a voiceless t. A phone menu line such as "Sipariş durumu için 1'e, iade için 2'ye basın" (press 1 for order status, 2 for returns) has to be read bire and ikiye. The voice handles this when the writer has chosen the right suffix; if the writer got it wrong, the audio will be wrong too.

Suffixes on names and foreign words

Proper names also take suffixes after an apostrophe: İstanbul'da (in Istanbul), Ankara'ya (to Ankara). For foreign names the suffix follows the original pronunciation: Facebook'ta, with -ta because the name ends in a k sound with a back vowel. Brand names are a common source of errors, both in writing and in reading.

Long words

Suffixes stack. Evlerinizden is ev-ler-iniz-den, "from your houses". Gelemeyeceklermiş is gel-eme-yecek-ler-miş, "apparently they won't be able to come". The famous showpiece Çekoslovakyalılaştıramadıklarımızdanmışsınız is a joke word, but it shows why a voice has to parse a word into its parts to stress and pace it properly.

Where does stress fall in Turkish words?

The default is easy: stress falls on the last syllable, and it moves as suffixes are added. Kitap, kitaplar, kitaplarım (my books) are all stressed at the end. The exceptions are what make a voice sound native or not.

• Place names often have stress earlier: Ánkara, İstánbul. Bebek (baby) is stressed at the end, but Bebek, the Istanbul neighbourhood, is stressed on the first syllable.
• Some adverbs are stressed on the first syllable: şímdi (now).
• Unstressed suffixes push the stress onto the syllable before them. The negative -me/-ma is one: gelmedi (he didn't come) is GEL-me-di. This creates real pairs: gélme means "don't come!", while gelmé means "coming", as a noun.
• The question particle mi is never stressed; the stress goes on the word before it.

Modern voices handle common words correctly. Mistakes appear on brand names, rare place names and foreign words. Fix those in the pronunciation dictionary; see the pronunciation guide.

What other details trip up Turkish voices?

• Soft g (ğ) isn't pronounced as a consonant; it lengthens the vowel before it. Dağ (mountain) sounds like "daa", yağmur (rain) like "yaamur", değil (not) like "deyil" or "deel".
• The circumflex separates words but is often left out: hala is a paternal aunt, hâlâ means "still"; kar is snow, kâr is profit. In a financial text, write kâr.
• Numbers and money: Turkish writes 1.250,75 TL with a dot for thousands and a comma for decimals: bin iki yüz elli lira yetmiş beş kuruş. A dot after a number makes it an ordinal: 3. is üçüncü (third).
• Abbreviations: vb. (ve benzeri, and similar), vs. (vesaire, etc.), Dr. and TL. If one is read wrongly, write it out.

How do you get natural Turkish text-to-speech?

1. Restore Turkish characters. Search for c, s, g, o, u and i in suspicious words and correct them.
2. Avoid all-caps text, or check every I and İ if you can't.
3. Pick a voice. Listen to samples on the Turkish voices page and choose one that fits the content.
4. Split long sentences. Turkish puts the verb at the end, so a long sentence such as "Geçen hafta İzmir'de açtığımız mağazamıza gelen tüm müşterilerimize teşekkür ederiz" (we thank all our customers who visited the store we opened in İzmir last week) makes the listener wait for the meaning. For audio, shorter is clearer.
5. Generate and listen line by line. Check names, numbers with suffixes and words with ğ. In the Leelo studio you can regenerate a single line and keep several takes.
6. Direct the delivery. Add pauses between menu options and slow down prompts for phone lines. See directing the delivery.
7. Export as MP3 or WAV.

Turkish voiceover suits phone menus, YouTube videos and training. To reach Turkish viewers with an existing video, dubbing translates and voices it in Turkish. A 1,500-word training script costs 1,500 credits in a standard voice or 2,250 in a premium one; plans are listed on the pricing page.

Frequently asked questions

Can Turkish text-to-speech read text without Turkish characters?

It will read it, but as written. "Cok" becomes "jok" and "sikinti" loses its ı sounds. Restore ç, ş, ğ, ı, ö and ü before generating audio; it takes a few minutes and avoids most errors.

Why does the voice say "Istanbul" wrongly in my headings?

Headings in capitals often lose the distinction between I and İ, especially after passing through software that uses English case rules. Write headings in sentence case, or check that İstanbul has its dotted capital İ.

How do I fix stress on a place name or brand name?

Add the name to the pronunciation dictionary with a spelling that produces the stress you want, then regenerate the affected lines. The setting then applies everywhere the name appears in the project.

Does Turkish text-to-speech handle vowel harmony?

Yes, in the sense that correctly written Turkish already shows the harmonized vowels, and the voice reads them. The weak point is suffixes on numbers, abbreviations and foreign names, where the writer must choose the suffix that matches how the word is said.

How much does Turkish voiceover cost?

In Leelo, 1 credit per word in a standard voice and 1.5 credits in a premium voice. Dubbing a video into Turkish costs 5,000 credits per minute. The free plan's 1,000 credits are enough to test voices on your own script.