AUDIO STUDIOG / P / L

Text to Speech

Turn your words into a voice.

31 languages · No sign-up · Text stays in your browser

Start hereIn your browser · No file uploads

Enter text → Generate → Listen & save

Choose a voice style

For the content you are creating

Presets choose a voice and speaking speed. Emotional acting is not supported by this model.

Select your text’s language. Site navigation and speech language can be different.

Your text

0 / 2,000 charactersPlain text only

Choose the language that matches your text. Up to 2,000 characters and three minutes of audio. Start with a short sentence on phones.

I agree to the model license and its use restrictions.

First generation downloads about 215 MB of model files and runtime. No download starts until you select Generate voice. Models are cached when browser storage allows; changing language uses the same models.

Ready when you are.

Voice settings

Voice Speed 1.00x

SlowerFaster

Video Narration

Use the preview to check pronunciation and choose the voice that suits your content. Very short text at high speed can fail; try normal speed or a full sentence.

How to use it

Choose the speech language, enter text and choose a voice or preset.
Agree to the model use terms, then select Generate voice. Keep this tab open while the first download and generation finish.
Listen to the generated audio, then download MP3 or WAV.

Before you start

Generation runs on your device. Mobile performance depends on the browser and available memory. Cancel stops processing immediately. Processing stops after five minutes.

This model supports the 31 languages in the selector, including Korean, Japanese, Spanish, French, German, Arabic and Hindi. Chinese is not supported. Use one selected language per generation; mixed-language pronunciation may vary.

Presets adjust the selected voice and speed; they do not rewrite text or add emotional acting. Names, abbreviations and numbers may be pronounced differently than expected. Listen before using your audio.

WAV is 44.1 kHz, 16-bit mono. MP3 is a smaller, lossy download encoded in this browser. There is no voice cloning. Closing this page removes your text and generated preview from this tool.

Privacy, model and open-source licenses

Text and generated audio stay in your browser. Only fixed public model and runtime files are downloaded; no cloud speech API or speech analytics is used.

Speech: Supertonic 3 by Supertone, with a community float16 conversion (q3146dq4/supertonic-3-fp16). Model: OpenRAIL-M, including use restrictions. This is a modified model, not an official Supertone service. Generated audio is synthetic; disclose that it is AI-generated when sharing it, as required by the model license.

Adapted inference helper: MIT. ONNX Runtime 1.26.0: MIT. MP3: wasm-media-encoders 0.7.0 (MIT) and LAME (LGPL 2.0 or later).

Word Counter · All tools

Using this tool

Select the language of your text, choose a voice and generate speech. Listen to the preview to check pronunciation before downloading MP3 or WAV.

What to check

The first generation downloads about 215MB of model files and runtime. Presets adjust voice and speed; emotional acting is not supported. Review the model license and use restrictions.