Text to speech
Paste your text, pick a language and voice, listen, then download the speech as an MP3 or WAV file.
0 / 300 characters
The audio is synthetic speech made by an AI voice. Do not use it to impersonate a real person or to mislead listeners.
Uploaded and processed audio is stored on SoundHalo. Audio access expires after 30 days; expiry does not confirm storage deletion. Delete the stored files here before leaving this page. Speech providers have their own retention policies. Privacy policy.
A free account reads up to 3,000 characters at a time and 20,000 a day. Pro reads up to 20,000 at a time and 300,000 a day. Sign in free
Example output
This file was made with this tool from the text below, using the Heart voice (English, US). It is 274 characters long and plays for 16 seconds. Nothing was edited afterwards.
Welcome to the weekly product update. This week we fixed three bugs in the export menu, and uploads are now about twenty percent faster.
If you have questions, reply to this email or join the live session on Thursday at four PM. Thanks for listening, and see you next week. How long text is read
The text is split at sentence ends into pieces of up to 1,000 characters. Each piece is generated separately, the silence each voice adds at the start and end is trimmed, and the pieces are joined with the same short pause you hear between sentences. A blank line between paragraphs gives a slightly longer pause. The finished file is set to a steady speech loudness (about -16 LUFS), so every voice plays at a similar level. Pasted emoji and control characters are removed before counting, so the counter shows exactly what you are charged for.
Check it before you download
- Listen to the whole file. A voice can stress the wrong word or misread a name.
- Write numbers, dates, units and acronyms the way they should be spoken.
- Use commas and full stops to control pauses. Very long sentences sound rushed.
- Make sure the language you picked matches the text. A Spanish voice reading English text sounds wrong.
Limits
- No account: 300 characters per request and 1,200 a day. This allowance is shared by everyone on your network, and all anonymous visitors share a daily SoundHalo budget, so it can run out early.
- Free account: 3,000 characters per request and 20,000 a day.
- Pro: 20,000 characters per request (roughly 20 minutes of speech) and 300,000 a day.
Characters include spaces and punctuation. Daily limits reset at midnight UTC. A request that reached the voice provider counts even if it fails.
Languages and voices
Lemonfox voices cover English (US), English (UK), Spanish, French, Italian, Portuguese (Brazil), Japanese, Chinese (Mandarin), Hindi. Lemonfox marks its non-English voices as beta. Giggy voices add German, Portuguese (Portugal), Russian, Turkish, Polish, Dutch, Swedish, Danish, Norwegian, Finnish, Romanian, Hungarian, Czech, Indonesian, Vietnamese, Filipino, Arabic, plus Spanish voices with a Spain accent and two more French voices. We kept only voices that a speech recognizer could transcribe back almost word for word from their sample, and dropped ones that read the wrong language.
Synthetic voices and acceptable use
Every voice here is synthetic. Tell your audience when they are hearing an AI voice if it could matter to them. Do not use this tool to imitate a real person, to make it sound like someone said something they did not, or for scams and harassment. Downloaded files carry a metadata comment that says they are synthetic speech.
Privacy
Your text goes to SoundHalo and then to one voice provider, Lemonfox or Giggy, depending on the voice. SoundHalo stores the generated audio and does not keep the text after the request. Provider copies follow their own policies. Read the privacy policy.
Related tools
FAQ
Is this text to speech free? +
Yes, for short text. Without an account you can generate up to 300 characters at a time and 1,200 a day per network. A free account raises that to 3,000 at a time and 20,000 a day. Pro allows 20,000 at a time and 300,000 a day. Voice samples are free to play and do not count.
Which company processes my text? +
SoundHalo sends the text to one voice provider. Lemonfox makes the English, Italian, Brazilian Portuguese, Japanese, Chinese and Hindi voices, the Spanish voices Dora, Alex and Noel, and the French voice Siwis. Giggy makes every other voice. SoundHalo stores the finished audio file, with access ending after 30 days. Do not paste confidential or personal information you would not share with these providers.
Can I use the audio in a video or a product? +
You keep the rights to the text you submit and the audio made from it, as described in the SoundHalo terms. Do not use the voices to impersonate a real person, to put words in someone else's mouth or to mislead listeners. Some platforms ask you to label realistic AI voices; follow their rules.
Can it clone my voice or a celebrity's voice? +
This tool uses the listed stock voices. With a free account you can clone your own voice, after reading a consent sentence, in the separate voice cloning tool (soundhalo.com/tools/voice-cloning/). Celebrities, public figures and anyone who has not given written permission cannot be cloned.
Why did it read a number or abbreviation wrong? +
Voices guess how to say numbers, units and acronyms. In our checks, the Japanese and Chinese voices did not say "MP3" clearly. Write numbers and abbreviations the way you want them spoken, for example "four PM" or "M P three", then generate again.
Can I upload a PDF or Word document? +
Open a plain text (.txt) file, or copy the text from your document and paste it. The tool reads text only, so remove page numbers, headers and footnotes you do not want read aloud.
Can I change the speed or pitch? +
Not in this tool. The players here have a playback speed control for listening, but the downloaded file keeps the voice's normal speed.