Turn up to 20,000 characters of text into natural AI speech on this device, then download it as WAV or Opus. Free, no signup.
What is this tool?
Turn text into natural AI speech on this device. Text to Voice runs the Kokoro-82M model in your browser with 18 American and British voices, nine speaking styles, and speed, pitch and volume controls. Long text up to 20,000 characters is read sentence by sentence, so nothing is cut off, and each sentence is highlighted as it plays. You can import .txt or .pdf files, add exact pauses, and download WAV or Opus files labelled as AI-generated. Nothing is uploaded.
How to Use
- Type or paste your text, or import a .txt or .pdf file (up to 20,000 characters).
- Pick a speaking style and voice, then adjust speed, volume and pitch if you like.
- Click Generate speech (Ctrl+Enter). Long text is read sentence by sentence, with progress shown as part 1 of N.
- Listen while each sentence is highlighted, then download the result as WAV or as a smaller Opus file.
Frequently Asked Questions
Does my text leave my device?
No. The Kokoro AI voice model (about 80 MB) is downloaded once and cached; speech is generated in your browser and the text is never uploaded.
Is there a length limit?
Up to 20,000 characters per run. The text is split into sentences and each one is generated separately, so nothing past the model’s per-call limit is cut off. You can stop part-way and keep the parts already done.
How do I add pauses?
Use [pause] for a short beat, [long pause] for a longer one, or an exact length such as [[break=800ms]] or [[break=2s]]. The Insert pause buttons add these at the cursor. A blank line adds a paragraph pause.
Which download formats are offered?
WAV works everywhere. Opus is several times smaller and is offered where the browser can encode it (Chrome, Edge and other browsers with WebCodecs). MP3 is not offered because the available browser MP3 encoders are not under a permissive license. Both downloads are labelled as AI-generated.
What if the AI voice doesn't load?
The page falls back to your browser’s built-in voice, choosing the closest accent and gender to the voice you picked and applying your speed, pitch and volume. That fallback can play but cannot be downloaded.
Other tools
There are more calculators, converters and editors on Wivrix. If something is missing, tell us.