Turn a new paragraph into a reading guided by your own voice sample. Start with a newsletter update, course introduction or spoken note.
Guests can generate up to 500 text characters per day. No sign-up or credit card required. See free and paid allowances.
Free audio is for personal, non-commercial use. Commercial use is included on all paid plans, including Starter, subject to the Terms and your rights to the content and voice.
Listen before you create
Hear new newsletter, course and podcast lines generated from an authorized synthetic reference. These examples are not recordings of your own voice. Listening uses no credits.
English · instant clone example
New words generated from the Anna reference below. Reference and result are synthetic speech.
“Hi everyone, this week we finished the first design. Next, we will test it with a small group.”
Original reference · Anna
“Welcome to the channel. Today we are turning a short idea into a clear voiceover.”
English · instant clone example
New words generated from the Anna reference below. Reference and result are synthetic speech.
“Welcome to lesson one. Start with the example on your screen, then pause and try it yourself.”
Original reference · Anna
“Welcome to the channel. Today we are turning a short idea into a clear voiceover.”
English · instant clone example
New words generated from the Anna reference below. Reference and result are synthetic speech.
“You are listening to Small Ideas, a show about the useful things we learn along the way.”
Original reference · Anna
“Welcome to the channel. Today we are turning a short idea into a clear voiceover.”
The example voices are synthetic. Your own result depends on your reference, automatic recognition and the new script; listen for both similarity and correct words.
Record or upload 5–20 seconds of your voice (about 10 seconds works best).
01
Text to speech in your own voice starts with two pieces of material. The reference is a short recording that shows how your voice sounds. The script contains the new words you want to hear. The words in the reference are recognized automatically when you generate, so you do not have to type them again. An optional correction field is available for names or unusual words; it is not a box for acting instructions. Keeping these inputs separate matters when you are adapting a newsletter or lesson, because the reference does not need to contain the subject of the new paragraph. Use a clean, ordinary sentence as the reference and put the new message in the script field. The generated reading is synthetic and can differ from your natural delivery.
02
Record or upload 5–20 seconds of your voice (about 10 seconds works best). Choose one speaker, a quiet room and a steady microphone position. Avoid a sample with music, multiple takes or a long silent opening. Listen back before submitting it. Automatic recognition uses the selected clip, and you can correct its words if needed. Keep a local copy of the recording for future sessions. This instant workflow requires the reference for each request; it does not automatically create a saved My Voice profile. Reusing a clear authorized reference can make your production process more consistent, but it cannot guarantee an identical delivery every time. If a sentence sounds different, compare the script and pronunciation before changing your entire reference setup.
03
A written newsletter often relies on headings, links and visual scanning. A spoken version needs to carry its context in the sentences themselves. Open with the subject of the update, introduce one change at a time and replace a bare link with a useful description of where to find the resource. Remove email footers, navigation labels and long tracking addresses before generating. The prefilled paragraph on this page is a writing example you can replace; it is not automatically sent anywhere. If the update contains personal details or an internal announcement, review those words before submitting them. End with a clear next step that still makes sense to someone who hears the clip without reading the original email.
04
The current script limit is 500 characters per request. Plan a longer lesson as short sections with a complete idea in each one. A section might introduce a task, explain a step or recap a result. Do not cut a sentence at the limit just to fit it: the clipped thought will be harder to understand and to join later. Keep headings and filenames in an external production document so you can order the downloaded clips. This tool does not automatically divide a long document, assemble multiple files or align audio with slides. After generation, listen to neighboring clips together and adjust gaps in your editor. A little room between ideas often matters more than squeezing every possible word into a single take.
05
Use punctuation to clarify the structure of the message. A comma can suggest a small pause, and a full stop gives an idea a clear ending, but neither is an exact timing command. Spell unfamiliar initials in a way that makes their intended reading obvious. Write a date or price as you would say it if the compact written form is ambiguous. When testing a name, use it inside a representative sentence rather than generating it alone. There is no pronunciation dictionary, word-level timing editor or guaranteed emotion control here. Bracketed directions such as whisper or smile may be read as ordinary words. For an exact performance, record the line yourself and use the recording workflow instead of expecting a punctuation trick to direct the model.
06
Listen for two separate things: whether the new words are accurate, and whether the voice is suitable for your project. Similarity alone is not enough if the reading changes a number or leaves out an ending. Check the first and last words, specialist terms and any change of tone between sentences. Download your MP3 while the temporary playback link is available and keep the original script beside it. For a video, add captions from the final audio rather than assuming the draft and recording are perfectly aligned. A clone is not proof that the speaker approved a message. If you use another person's authorized voice, follow the scope of that permission and do not present synthetic speech as an authentic recording of them.
07
Guests share 500 daily characters between standard speech and instant cloning. A signed-in account uses its displayed credits. Existing example playback does not generate a new recording or spend that allowance. Before making another take, decide what you are changing: a difficult name, an overlong sentence or an unclear transition. Revise one issue at a time so you can hear whether the change helps. Avoid repeatedly generating an unchanged paragraph in the hope that every problem will disappear by chance. A failed request returns its reserved user allowance, while a completed take that you prefer less is still a completed generation. Keep the useful versions and assemble the final spoken update in your usual editor.
A little clarity
The details that help you create with confidence.
Contact us →Guests can use 500 characters a day, shared between standard speech and instant cloning. Signed-in users use their displayed credits. Listening to existing samples uses no credits.
Record or upload 5–20 seconds of your voice (about 10 seconds works best). The reference words are recognized automatically. Correct the transcript if needed, then put the new words in the separate script field.
The instant workflow uses your reference for each request. Sample words are recognized automatically, and the current recording can be reused while the page stays open. Signing in does not create a permanently saved voice profile.
The current request limit is 500 characters. Divide a longer newsletter or lesson into meaningful sections yourself; the tool does not automatically split and join a document.
You can edit the spelling and punctuation of the script before generating. There is no word-level timing editor, pronunciation dictionary or exact pause-duration control, so check names and specialist words in the result.