Clean the text, spell numbers out in words, split it into sentences and see exactly how many text-to-speech requests it will take — all inside your browser. Nothing is uploaded, nothing is sent anywhere, and it keeps working with the network switched off.
Speech engines guess at bare digits and often get them wrong, which is
why 1,180 can come out as something that is not a number at all. Writing
numbers out removes the guess. Thousands separators, decimals and currency symbols are
all handled, and codes attached to letters like MP3 are left alone.
Separator rules follow the language you pick, because they are the
opposite of each other: English reads 1,180 as thousands and
3.14 as a decimal, Turkish reads 1.180 as thousands and
3,14 as a decimal. Turkish also drops the apostrophe when a suffix joins a
written-out number.
Every button changes the text in the box above, so you can run them in any order. There is no undo — copy your original first if you need it.
Sentences are split on . !
? Click any segment to copy it.
Paste some text to see the segments.
Speech services bill by the character, so the figure that matters is the readable character count below. Divide it into your own plan's allowance to see what the text costs you — providers price differently and change their rates, so this page deliberately does not guess a price in money.
Readability follows the language you selected. English text is scored with Flesch Reading Ease; Turkish text is scored with the Ateşman index, the Turkish adaptation of the same formula. Applying the English formula to Turkish (or the reverse) produces a meaningless number, so the two are never mixed. Syllables are counted by vowel groups in English and by vowels in Turkish, which is how each language works.
Reading time is words divided by a speaking rate. 180 words per minute is a typical narration pace; audiobook narration usually runs slower and podcast conversation faster. Treat it as a bracket, not a stopwatch.
Nothing leaves your browser. There is no server call on this page, no analytics and no tracking. The optional history is stored only in this browser's localStorage and is off unless you switch it on; clearing it removes it immediately.
Speech engines guess at bare numbers, and a wrong guess is impossible to spot until you listen to the whole file. Writing the number out in words removes the guess entirely.
Most engines cap a single request at around 5,000 characters. Knowing the split in advance stops a long chapter failing halfway through.
URLs, e-mail addresses and repeated whitespace are billed like any other character and are read aloud as gibberish.
Knowing a chapter runs 42 minutes at a normal pace tells you whether to split it before you spend anything on narration.
No. Every calculation on this page runs in your browser. There is no server call, no analytics and no tracking, and the page keeps working with the network switched off.
By the language you select, because the two conventions are opposites. English reads
1,180 as one thousand one hundred eighty and 3.14 as a
decimal; Turkish reads 1.180 as thousands and 3,14 as a
decimal. Picking the wrong one would turn a price into a completely different
number.
English reads them digit by digit, so 3.14 becomes
three point one four. Turkish reads the fraction as a number, so
31,90 becomes otuz bir virgül doksan. Each follows the
convention of its own language.
No. Digits attached to letters are left alone deliberately, so product names, codec names and model numbers survive.
Flesch Reading Ease for English and the Ateşman index for Turkish. They are the same idea calibrated for different languages; applying one to the other returns a meaningless figure, so the page never mixes them.
Only if you switch history on, and then only in this browser's own storage. It is off by default and “Forget all” erases it immediately.
This page belongs to Bubixo Voice and Speech Tools, where every tool in the category is listed. These are the ones people usually open next.