Free · Windows 10 / 11 · No install · No account
Voice Maker
Need a voice-over without recording one, or a narrator for a language you don't speak well? Voice Maker turns text into speech on your own computer, in 31 languages, with ten female and male voices. Listen, then save it as WAV or MP3. No account, no character limit, and the text you type never leaves your PC.
What it does
- 31 languages. English, Korean, Japanese, Vietnamese, Indonesian, Hindi, Arabic, Spanish, French, German, Italian, Portuguese, Russian and more European languages.
- Ten voices. Five female, five male, at 0.7 to 1.5 times their own pace, in three levels of quality.
- Long texts. Up to 20,000 characters in one go: the program reads a few sentences at a time and joins them for you.
- Sounds between the words. Buttons put a laugh, a breath or a sigh where you want it in the text.
- Listen first, then save. The player is in the window; save what you like as WAV or MP3 into Music\Voice Maker. Nothing is ever written over.
- On your computer. After the one-time model download, speech is made with no internet connection, on the processor alone: no graphics card needed.
- No account, no limits. No sign-up, no daily character quota, no audio watermark.
- Your text stays with you. What you type and the speech made are sent nowhere, and the text is not written to any file.
Download Voice Maker
- Version
- 1.0.0
- Date
- 19 Nov 2026
- Size
- 14.5 MB
- SHA-256
- 06b246a09413ed40371be7ab54ea9c9c0862277fb67769891b035346543fe7d0
To check the file hasn't been tampered with, compare the SHA-256 above with the output of Get-FileHash VoiceMaker-1.0.0.zip in PowerShell.
Getting started
- Download the zip above, right-click it → Extract All… into a folder of your own, such as Documents. Don't run it from inside the zip, and don't keep it in Program Files. The folder holds VoiceMaker.exe and onnxruntime.dll: the two must stay together.
- Double-click VoiceMaker.exe. The first time, a notice about usage statistics appears: read it and click Got it.
- Read The model's terms of use, tick I have read and accept the model's terms of use and press ⬇ Download the model (about 383 MB, once; you can stop and go on later).
- Type or paste the text, set Language of the text to the language it is written in, choose a Voice and press 🔊 Make the speech.
- Listen in the window, then press 💾 Save as WAV or 💾 Save as MP3. 📂 Open the folder shows the file.
The first time you run it, Windows may show a blue "Windows protected your PC" screen because the program is not signed with a paid certificate. That doesn't mean it has a virus. If you downloaded it from this site, click More info (the link under the warning), then the Run anyway button that appears (once).
Good to know
- When you publish the speech, say clearly that it is AI-made, and never use it to impersonate anyone, or to deceive, defame or harass others. These are conditions of the model's licence (OpenRAIL-M). The platform you post on may have its own rules about AI voices.
- The ten voices are ready-made. The program cannot copy a person's voice.
- Only the 31 languages listed. A text in another language (Thai or Chinese, say) is not read: the program says so rather than making noise.
- For a good reading: choose the language the text is written in, spell out numbers and abbreviations that come out wrong, use full stops and commas where you want a pause, and an empty line for a new part. Making it again gives a slightly different reading each time.
- Speed depends on the computer. On an 8-core laptop, 8 seconds of speech take about 3 seconds to make; a slower machine takes longer. While making speech the program uses the whole processor and about 1 GB of memory.
- The voice model is outside the program's folder, in %LOCALAPPDATA%\VoiceMaker\model (about 383 MB), so updating the program does not lose it. ⚙️ Settings → Voice model shows where it is, opens the folder and can delete it. On another computer the model has to be downloaded again.
- MP3 is written by the encoder that comes with Windows (192 kbps). An N edition of Windows without the Media Feature Pack offers WAV only.
- The speech stays in memory until the next one is made or the program closes. Press save to keep it.
Requirements
- 64-bit Windows 10 or Windows 11. Once the model is downloaded, speech is made with no internet connection (the internet is used only for the one-time model download, ads, usage statistics and updates). 8 GB of memory or more is recommended, and about 500 MB of free disk space.
- The program's window uses WebView2, which comes with Windows (without it, the program opens in your default browser instead).
- The program is in English or Thai: it picks the language from Windows and you can switch under ⚙️ Settings → ภาษา / Language.
Ads and removing them
Every feature of Voice Maker is free. It shows a small ad bar at the bottom of the window (one line of text, or a picture of 468×60, 728×90, 970×90 or 320×50, scaled down when the window is narrow) and, when the window is wide enough, an ad box beside the text box (text, or a picture of 300×250 or 300×600, scaled down to fit the window's height). To remove the ads, ask for a key at contact@freesabai.com, then click Remove ads on the bar, paste the key into Ad-removal key and click Enter key. Keys are checked on your computer, without the internet.
Privacy
The text you type and the speech made stay on your computer: nothing of them is sent anywhere, and the text is not written to any file (the program remembers only your last choices, such as language and voice). When you download the model the first time, the program contacts Hugging Face (huggingface.co) directly, at your click, so your IP address reaches it as with any download from a website; after that it is not contacted again. Apart from that, Voice Maker sends only a usage report at most once a day (a random ID for the installation that the program made itself, not taken from the hardware, the version, the Windows version, the Windows language and region, whether a key is in use, and how often each ad was shown). The first time it opens, the window shows a notice about this before anything is sent. Switch it off by clicking Usage statistics: on under ⚙️ Settings → Privacy. You are still told about new versions, because the latest version number comes with the ads the program already downloads. When you click to update the program, it downloads the new version from freesabai.com and sends nothing else. The full details are in our privacy policy.
Updating and removing
- When there's a new version of Voice Maker, even with usage statistics off, a strip under the top of the window says the new version is out (Later hides the strip until the program is next opened and leaves an ⬆ Update to button with the new version number at the top of the window, next to ⚙️ Settings, for later). Click Update now on the strip, or that button, and the program asks first; click Update now and it downloads the new version from freesabai.com, checks our signature, then closes and starts again by itself. Your settings stay; other files in its folder are only replaced when they are Voice Maker's own, such as the README.txt from the zip. (Click See the download page if you'd rather download it yourself from the program's download page on freesabai.com.)
- To check yourself at any time, go to ⚙️ Settings → Check for a new version (the program asks freesabai.com only when you click this button). If the check fails (for example without internet), the program opens its download page on freesabai.com. The program never downloads a new version without your click.
- If it can't update itself (for example its folder can't be written to, or the download fails the checks), it opens the program's download page on freesabai.com instead. You can also close the program and unzip the new zip over the old folder; your settings stay.
- To remove: close the program and delete its folder. What it remembers about this computer (the language, your last choices, the ad-removal key, the random ID for the statistics and the voice model of about 383 MB) is in %LOCALAPPDATA%\VoiceMaker; you can delete that too (or press 🗑 Delete the model in ⚙️ Settings before deleting the program). The sound files you saved are in Music\Voice Maker and are not deleted.
Problems and ideas
Found a problem, need another language or voice, or want Voice Maker to do more? Tell us on the feedback page. No sign-up needed.
Questions and answers
Is there a free text to speech program for Windows that works offline?
Yes. Voice Maker turns text into speech on your own computer. After a one-time download of the voice model (about 383 MB), it needs no internet connection, no account and has no daily character limit. Type or paste the text, choose the language and one of ten voices, press Make the speech, then save it as WAV or MP3.
Which languages can Voice Maker read?
The 31 languages of the Supertonic 3 model: English, Korean, Japanese, Arabic, Bulgarian, Czech, Danish, German, Greek, Spanish, Estonian, Finnish, French, Hindi, Croatian, Hungarian, Indonesian, Italian, Lithuanian, Latvian, Dutch, Polish, Portuguese, Romanian, Russian, Slovak, Slovenian, Swedish, Turkish, Ukrainian and Vietnamese. Choose the language the text is written in; it cannot read a language outside this list.
Why does it download 383 MB the first time?
That is the voice model, too large to put in the zip. You press the button once and the program fetches it from Hugging Face, where the model's makers published it, checking every file before use. From then on speech is made offline. The model stays on your computer; ⚙️ Settings → Voice model shows where it is and can delete it.
Can I use the voices in monetised videos, and is my text sent anywhere?
The model's licence (OpenRAIL-M) does not forbid commercial use, and the model's makers claim no rights in the speech you make. It has conditions: when you publish the sound you must say clearly that it is AI-made, and you may not use it to impersonate anyone or to deceive. The platform you post on may have its own rules about AI voices. The text you type and the speech made stay on your computer and are sent nowhere.