English text to speech. English only, on purpose.

Many tools list dozens of languages and do all of them a little. Clonylab does one and tries to do it well. Every voice here is built and chosen for English, and the whole product, from the Studio to the clone check, is made around English text.

That has a cost, and we would rather be clear about it. If you need Spanish, Hindi, French or any other language, Clonylab is not the right tool. If your work is in English, you get a focused set of voices, US and UK accents, and your own cloned voice for English scripts.

The biggest choice is the accent. US and UK voices read the same spelling in different ways, and your audience hears the difference at once.

Hear them in groups on the British text to speech and American text to speech pages, or on the standard voices page.

Text that looks fine on a page can sound odd when read by a voice. These changes help in any English voice.

Spelling differences matter too. Write “colour” or “color” as you like, because the voice says the same word either way. What changes is the voice’s accent, so choose that first.

English has rhythm, and a script can set it. Add [pause 1s] where a speaker would stop, wrap important words in [slow] and [/slow], and use the speed and pitch sliders for the overall feel. All of this works in every English voice.

The AI voiceover generator page lists every tag.

• It is text to speech, not translation. Clonylab reads what you write. It does not translate.
• Mixed-language scripts sound mixed. A French phrase in an English script is read with an English accent.
• Names can need help. Respell unusual names the way they sound.
• The first use of a voice downloads it. Your browser keeps the files.

More about the basics on the text to speech online page.

Here is a short script, as someone might first write it, and the version that is ready for a voice.

• The abbreviation “Dr.” became “Doctor”.
• The date is spelled out so it is not read as a month and a day in the wrong order.
• The number is written as it should be said.
• The address is spelled the way people say it.
• Two small pauses make it easy to follow.

If your listeners are learning English, slower and clearer helps. Lower the Speed slider a little, raise the Sentence pause a little, and use [slow] on new words. Pick a voice with a clear, steady delivery and test a full paragraph.

These are starting points. Listen, change one thing, and listen again.

People speak at about 150 words a minute, so a quick estimate works well. Slower voices and pauses add time.

The Studio shows your characters and words as you type, so you can see where you are.

• Prefer short words. “Use” beats “utilise”.
• One idea per sentence. The listener cannot re-read.
• Say who does what. “You press Start” beats “Start is pressed”.
• Repeat key words instead of swapping synonyms. Listeners follow repetition easily.

• English only: 38 voices, 28 US and 10 UK.
• Write for the ear: short sentences, spelled-out numbers and respelled names.
• Direct the reading: pauses, [slow] and the sliders work in every voice.
• Test a short version before you generate the whole script.
• Start free: 300 characters, one time, no card.

• Choose by accent: British or American.
• Pick a voice: the AI voice generator page.
• Clone your own English voice: voice clone online.

English is full of words spelled the same and said differently. A voice has to guess from the sentence. Most of the time it guesses right. When it does not, you can help it.

Change one thing at a time and generate again. A respelled word takes a second to test.

• Read it aloud yourself once. Anywhere you stumble, the voice may stumble too.
• Spell out numbers and symbols the way they should sound.
• Respell unusual names and brand words.
• Add pauses where a speaker would breathe.
• Generate a short test before the full script.
• Listen to the whole thing before you share it.

The same habits help in every voice, US or UK, standard or cloned.

Questions about English text to speech