Any text you need, read aloud in the voice you choose.

Write your own text, or describe what you need, and Kukarella writes it for the people who have to hear it. Choose a voice in the accent you want, or let the assistant choose one. Press play, and download the audio as MP3 or WAV. You do not need an account. You do not need a credit card.

Every text you teach can have audio.

What voiceover text would you like to create?

You can also say your request out loud, or upload your own file and tell us what to make from it.

Type or paste your own text and choose your own voice.Type or paste your own text and pick the voice.

How you get from text to audio

Three steps. Writing the text is free. You pay only when you press play.

Step one: the text

Write it yourself, paste it, or upload a file. Or describe what you want and the assistant writes it, at the level and in the language you ask for.

Nothing is spent yet. Until you press play the text is only a document, and you can work on it as long as you like.

Step two: the voice

The assistant always starts you on an English voice. If your text is in another language, change it. If you wrote the text yourself, choose the voice yourself.

Samples are free, so nothing is spent while you are deciding.

Step three: press play

This is the only step that spends characters. One voice reads the whole text, and the audio is yours.

Give the same text to several voices and you get one file from each. Each voice reads the whole text, so each one costs characters again.

You can do all three steps without an account, and without a card. Nothing you make is charged until you press play. Three requests without an account; a free account gives you ten.

Try it with your own request

Have your own text?

Where your text can come from

Most people already have the text. The rest want help writing it. Either way, nothing has to be retyped.

Files

Upload a chapter, a worksheet, an article or your own notes. DOCX, PDF and TXT. The words come out as plain text, and you can change any of it before you press play.

Audio and video

MP3, WAV, M4A, AAC, FLAC, OGG, MP4, MOV, AVI, WEBM and MKV. Kukarella turns the speech into text. An old recording becomes text a new voice can read.

A web page

Paste the address of a page and you get its text. Useful when you want your class to hear an article instead of reading it. A news story, a blog post, a museum page — anything with text on it.

A photograph

JPG or PNG. A picture of a textbook page becomes text you can edit. You can also ask the assistant about a photograph. Ask who these people are, where they are, and what is happening. It writes a narration from the answer.

How much at once

Give the assistant up to 10 files, 20 MB each. Inside a project you can upload 3 files of up to one gigabyte. Making audio spends characters. Turning a recording into text spends minutes.

The Chrome extension

Sends text straight from ChatGPT, Gemini, Claude and Perplexity into a new project. Select the text, right-click, and a project opens with it already inside.

Listen to two more examples

Each of these came from one sentence, or from a text that already existed. Change the level. Change the language. Change the accent. It is the same amount of work.

The water cycle, for a primary class

English (United States) · native voice

0:24

One idea per sentence, in the order it really happens. This is what a narration for eleven-year-olds has to do.

Water is always moving. The sun warms the sea, and some of the water turns into a gas you cannot see. The gas rises. High in the sky it gets cold again, and it turns back into tiny drops. Those drops make clouds. When the drops get heavy, they fall as rain. The rain runs into rivers, and the rivers run back to the sea. Then it starts again.

B1 Spanish — a day at a local market

Mia S. · Spanish (Mexico) · native voice

0:26

Mexican Spanish, not "Spanish". Your students hear the accent real people use.

El mercado de la colonia abre todos los días a las seis de la mañana. A esa hora llegan los camiones cargados con la fruta y la verdura del campo. Doña Elena vende naranjas en el mismo puesto desde hace más de veinte años. Conoce por su nombre a casi todos los clientes que pasan por su pasillo. A mediodía el mercado está lleno de gente que busca algo para la comida. Hacia las cuatro de la tarde los puestos empiezan a cerrar poco a poco.

Everything you hand out is silent

Your learners hear only you. Your accent. Your speed. Your way of saying things.

The reading text, the article, the page you photographed from a book. Your students can read all of it. They can only listen if you read it aloud yourself, every time, to every group.

Your course book has audio, but it was recorded for a different class and a different syllabus. The level is right for some of your students and wrong for the rest.

Kukarella reads your material out loud. You choose whose voice does it.

The same text, five accents

A2 English — taking the train to work

I leave the house at half past seven every morning. The station is about ten minutes away, so I walk. I buy my ticket from the machine near the door. The queue at the ticket window is always very long. My train comes at eight o'clock and it is usually full. I find a seat near the door and I read. Then I get off and walk five minutes to my office.

Present simple only. Every sentence is at A2 level. Words an A2 group already knows. Written to be heard, not read.

The text you teach becomes listening practice

Paste it in, choose a voice, press play. Ask for the same page one level down and your Tuesday group gets its own version.

The accent your students will really hear

Your class will meet one particular accent, not a general one. Choose the country your syllabus points at, and they practise on the real thing.

A voice that is not yours, when it helps

Sometimes it helps that the voice is not yours. A machine reading a hard sentence does not sound like blame.

The lesson works when you are not there

A student who missed Tuesday can listen at home. A substitute teacher can play it. It does not need you in the room.

What you can say when somebody asks why you bought it

Every text we teach now has audio. I make it myself, in the accent I want, and I do not have to buy another course book.

Make the first one for your class

Have your own text?

Listen to every voice before you sign up

Many people come here for one thing: a voice with the accent they need. Kukarella has more voices than you will ever need. The point is finding that voice.

Kukarella has native voices for Canadian French and for Brazilian Portuguese. English comes in 14 accents and Spanish in 23. In total there are 2,000+ voices in 152 languages and accents.

You do not have to believe any of this. Open the library and listen.

Choose a language, then a country.
Spanish from Spain, Mexico, the United States and twenty more countries. English from the UK, the US, Australia and India. Open the list and see them all.
Listen to any voice.
Press play on any voice in the library and hear it read a sample.
Voices you cannot use yet show a lock.
Hover the lock and it says what opens the voice. Locked voices still play their sample, so you hear what a paid plan adds.
Native voices come first, multilingual voices after.
Which kind you pick changes how your Spanish lesson sounds. The next part explains why.

Found a voice you like? Paste your text and give it to that voice.

How to choose a voice, and why some sound wrong

Please read this part before you start. Choosing the wrong kind of voice is the most common reason people think an AI voice sounds bad.

There are two kinds of voice. They are good at different jobs.

Native voices

1,900+ of them

A native voice speaks one language only. It knows how that language really works: where the stress goes, how numbers are read, how a question sounds. It is clear and correct.

A native voice has a real regional accent, and you choose which one. Mexican, American, or Spanish from Spain. British, Australian or Indian English. A native voice cannot read another language.

Multilingual voices

about 300 of them

One multilingual voice reads many languages, not one. This is useful when the same text has to exist in several languages in one voice.

No multilingual voice is excellent in any one language. Listen to a whole paragraph before you decide, because mistakes in stress and pronunciation show up over time. Each one also reads a language with one fixed accent — usually the accent of its English recording, which is why Spanish can come out American-sounding.

Hear the difference on one text

All six voices read the same three sentences from the example further up this page.

Native — English (UK) only

Multilingual — reads any language

Listen to the vowels, and to where the stress falls. That is where a multilingual voice gives itself away.

So how do you choose?

One language, and the accent matters→use a native voice. For a language lesson this is almost always the right answer.

The same text in several languages, in one voice→use a multilingual voice, and accept the accent it comes with.

If your Spanish sounds a little American, or your Finnish has an English t and r in it, you are using a multilingual voice. Choose a native voice instead. The accent will be correct.

Cloned voices, and when they are worth it

You can add a voice the library does not have. It joins your voice list, and you pick it like any other. Cloning needs a paid plan.

From an audio file

Upload a recording of the voice. Up to 20 MB in total, and no more than five minutes of audio.

From a recording made here

Record straight into the browser. The same limits apply. No file to find and nothing to install.

From a description

Describe the voice you want and the AI builds one. No recording, and nobody real is copied.

Where a clone earns its place

A voice you cannot record. Somebody who is no longer here. A historical figure. A character you invented. Nobody in the room knows how they really sound, so nothing gives it away.

Where it lets you down

Your own voice, read to your own students. They know your timing and your habits. A clone gets the small things wrong — a pause in the wrong place, a word said your way but not quite — and your own students hear it. It does not fool them. It makes them trust the voice less.

What it costs to use

A cloned voice spends three characters for every character of your text. That is three times a normal voice, so keep it for the parts that need it.

Hear a clone next to the voice it came from

Press the original, then press the clone. The same words, twice.

Dorothy, who works with us

She recorded two sentences in the browser and gave us permission. Her clone reads the same two.

Theodore Roosevelt, 1912

The original is a campaign recording, and it is in the public domain. The clone reads words he never said.

We had the right to clone both of these. Clone a voice only when you have that right too.

The AI writes a draft. The project is yours.

Nothing the assistant makes is locked.

Change the voice

Swap the voice at any time, even after the audio exists. The text stays exactly as you wrote it, and only the reading changes.

Edit like a document

Change any word at any time. Or paste your own text at the start and never open the assistant at all.

Add pauses and sound

Ten pause lengths, from half a second to a minute and a half. There are also 2,900+ sound effects, and you can add your own audio.

Rewrite one paragraph

Ask the AI to reword it, simplify it or translate it. The rest of the text stays exactly where it was, and so does the voice.

Download what you need

MP3 or WAV, as one file or split by paragraph. Split files suit video editing. The text itself exports as TXT or DOC.

Share it with others

Send a link that plays without an account. Or invite up to 10 people by email as editors or viewers — they need a free account.

Say who is listening, and the text is written for them

Ask for A2 and the text is written in A2 language. Short sentences, and words an A2 class already knows. You do not have to name a level. You can say "for people learning the language". Or "for a first-year class". Or "for somebody new to the subject". CEFR, HSK, JLPT and TOPIK are just the most exact way to ask. They work in every language, not only in English. Ask for B1 German and you get German.

Levels

CEFR A1
A2
B1
B2
C1
C2
HSK
JLPT
TOPIK
"beginner"
"elementary"

Exams, in the language they are taken in

IELTSEnglish
TOEFL iBTEnglish
TOEICEnglish
DELESpanish
DELFFrench
GoetheGerman
HSKChinese
JLPTJapanese
TOPIKKorean

Any language, not only English

Ask in the language you teach and you get that language. Not the language of the file you uploaded. Not the language of your source text. If you asked for German, you get German.

Sentence length is set for each level and each language

Every language measures a sentence differently. Words for English and Spanish. Characters for Chinese and Japanese. Groups of words for Korean. A B1 text has shorter sentences than a C1 text.

Ask for a level now

Name the level and the language. That is the whole request.

A free account triples what you can try with

30,000 characters instead of 10,000, and 30 minutes of transcription instead of 10. Every voice in the library opens on any paid plan.

What you can make

Ten kinds of voiceover. You describe each one in a sentence.

Language practice

Any level from A1 to C2, in any language you teach. Say the level and the topic, and you get a narration your class can follow, with sentence length set to that level.

"Write a two-minute A2 Italian narration describing a house and the things in each room."

Exam practice

Name the exam and get a talk or an announcement shaped like its spoken sections. Each one works in the language that exam is taken in.

"Write an IELTS Listening Part 2 style talk introducing a new community sports centre and its opening times."

Lessons and courses

Any subject, any age group, any length, in any language. Photosynthesis for eleven-year-olds, compound interest for people who fear maths, or how bread is made.

"Write a three-minute narration explaining photosynthesis for eleven-year-olds."

From your own file

Attach a document or a photograph and say who is listening. The same source can come back as a C1 summary or as an A2 retelling.

"Using the attached text, write an A2 narration that explains it simply to a language learner."

Not all of this happens in a classroom. A school also has a phone line, announcements and an open day. So does every other kind of business.

Employee training

New staff, safety rules, procedures and policy. Record it once and every new starter hears the same thing, in the language they read best. For each language, pick one voice in the accent your staff hear every day. Everything you make then sounds like one person. Upload the policy you already have and it comes back read aloud.

Phone systems

The message on your office line. Out of hours, holidays, a snow day. Phone systems need an 8 kHz WAV. You can download one: u-law for the Americas and Japan, A-law for Europe and elsewhere. You change it yourself, in minutes.

Ads and promos

An open day, a new course, a notice for parents. A radio spot or an in-store announcement. Say how long it should be, then check the length of the audio. Any language, and the accent your listeners really speak.

Demos and tutorials

How to use the school portal, or how to use your product. Write it once, then make the same walk-through again in another language, without booking a studio.

Video and social

Voiceover for video, reels and shorts, without recording yourself. Download one file per paragraph, and put each file on its own clip in your video editor.

Meditation and wellness

Calm reading for form time, for exam nerves, or for a wellness channel. You choose the pause between paragraphs, up to a minute and a half. The silence is already in the file.

Money, and what happens if you stop

These are the four questions people ask before they pay. The answers do not change.

If I stop paying, do I lose my work?

No. Your projects stay, and everything you have already turned into audio stays with them — you can still play it and download it. Files you uploaded can be removed if you use more storage than your plan allows.

A paid plan lets you keep changing things: the full voice library, and new credits every month. If you come back later, your projects are where you left them.

What if it is not what I expected?

Ask for your money back within 15 days of your first payment. Write to support@kukarella.com. Cancelling is different, and you can do it in your profile at any time.

Your plan then runs to the end of the period you paid for. If you cancel and come back within 30 days, your unused credits are still there.

Can I use the audio in a course I sell?

Yes, on a paid plan. Commercial rights to every AI voice are included, for courses you sell and for client work — as long as you hold the rights to the text you turn into speech.

The free plan does not include them. It is there for trying the product and learning how it works before you decide.

I only need this once.

Then take a plan for a single month. For one project that is almost always the cheaper route. Here is what each way of paying actually gives you.

See plans and top-ups

One month of a plan

Prime Lite, Prime or Prime Plus. That month's credits, every voice in the library, and commercial rights for the audio you make. A plan does not stop by itself: if you need the one month only, cancel it before the month ends. You can do that yourself, at any time.

A year of a plan

About a third cheaper than paying month by month, and every monthly allowance multiplied by twelve. This is the one to take if you will be making material all year.

A top-up

For a different case: your plan is right, and you just need a few more credits this month. Bought once, never expires. It does not open any voices, so it is not a way to skip a plan.

Questions people really ask

Taken from our own mailbox, not invented. The answers here are the answers we give.

Can I try it without an account?

Yes. Paste your text or describe what you need, choose a voice, and press play. You need a free account to save your work, and a paid plan for the rest of the voice library.

Can I use my own text, or does the AI have to write it?

Your own text. Make a project, paste it in, choose a voice and press play. Many people arrive with the text already written. The assistant is there for when you do not have it yet.

Can it read my language?

Almost certainly. Kukarella covers 152 languages and accents. Dutch, Turkish, Finnish, Vietnamese, Arabic, Japanese, Korean and Chinese are all there. Arabic, Chinese and Dutch come in several accents; for most languages there is one.

Why did my Spanish sound American?

You were using a multilingual voice. It reads a language with one fixed accent, and often with its English recording — which is why your Spanish came out American-sounding. Change the voice to a native Spanish one and choose the country you want. The section above explains why the two kinds sound so different.

Can I turn a document, a video or a photo into audio?

Yes. Use the Upload files button. It reads DOCX, PDF and TXT files. It reads audio and video files. It also takes text from a web page, or from a photo of a page.

How long is the text allowed to be?

There is no limit on one text. You have a number of characters and you spend them. A guest starts with 10,000. Most voices spend one character for each character of your text. A cloned voice spends three.

What level can it write at?

CEFR A1 to C2, HSK 1 to 9, JLPT N5 to N1, TOPIK level 1 to level 6, or simple words like "beginner". For exams: IELTS, TOEFL iBT, TOEIC, DELE, DELF, Goethe, HSK, JLPT and TOPIK.

Do I pay for a text the AI writes?

No. Writing a text costs nothing. Characters are spent only when text becomes audio.

If I change one sentence, do I pay for the whole text again?

No. Only the paragraph you changed is made again.

What can I download?

Audio as MP3 or WAV. Split paragraph by paragraph if you are dropping it into a video. The text exports as TXT or DOC, and for phone systems the WAV also comes at 8 kHz in u-law or A-law.

Do you have an API?

No. Kukarella has no public or developer API, and we are not planning one at the moment. Everything happens inside the app.

What do I get for making a free account?

More credits to try with. 30,000 characters for making audio, instead of 10,000. 30 minutes for turning recordings into text, instead of 10. 18 image credits instead of 8. Your work is saved, and it stays saved.

I only need this once. Do I have to subscribe?

Only for a month. For one project the cheapest route is usually a single month of a plan. You get that month's credits and every voice in the library. A plan does not stop by itself, so cancel it before the month ends if you need one month only. A top-up is for something else: your plan is right and you need a few more credits. A top-up does not open any voices.

Which voice does the assistant choose?

One of the same fixed set of 22 voices, for everyone and on every plan. The language of your text does not change that set, so a Spanish narration can open on an English voice. Change it yourself — that takes seconds, and the whole library is open to browse first.

We chose one voice for its accent and use it for all our audio. Will it still be there next year?

Every voice in the library is made by another company. If that company stops offering the voice, it can leave our library too. This is rare. If it happens, send us a few recordings you made with that voice. We can make a clone of it and add it to the library for everyone. The clone sounds very close to the original, and it will most likely read more than ten languages.

Can I use my own voice?

You can clone it on a paid plan, from a recording or from a description. Read the section above first. Students who know your voice hear the difference. A clone of you often works less well than a voice that is plainly not you.

Something is not working. Who do I ask?

Ask us directly. The chat is the message icon in the bottom-right corner of the site. Or write to support@kukarella.com.

Now describe what you need

Or bring your own text and pick the voice yourself.

Free to try, and the audio you make is yours to keep.