Voxtonic - Offline AI TTS

Voxtonic is an Android app that converts text into spoken audio. It runs the Supertonic 3 neural text-to-speech model directly on your device through ONNX Runtime, so synthesis works without an active internet connection after the initial model download.

Supertonic 3 is an open-source on-device TTS model released by Supertone on Hugging Face. Voxtonic packages that model into a mobile interface with a script editor, voice picker, library, and prompt templates. Voxtonic is an independent project built by an indie developer and is not affiliated with or endorsed by Supertone.


What you can do

Type or paste a script into the editor and tap Generate to produce a WAV audio file. Replay the result inside the app, mark it as a favorite, or remove it from the Library tab.

Each generated WAV file can be saved to your device or shared with other apps through the system share sheet, so you can move it into your audio editor, messaging app, or cloud storage of choice.

The app includes ten preset voices — five female (Aria, Luna, Nova, Iris, Echo) and five male (Atlas, Onyx, Cipher, Vega, Rune).

Synthesis is available in 31 languages, including English, Korean, Japanese, German, Spanish, French, Italian, Hindi, Indonesian, Arabic, and others such as Bulgarian, Czech, Danish, Greek, Estonian, Finnish, Croatian, Hungarian, Lithuanian, and Latvian.

You can insert three expression tags directly in your script — laugh, breath, and sigh — by tapping the chips above the text field. These tags are interpreted by the Supertonic 3 model during synthesis.

A small library of prompt templates is included to help you get started.

The Voice Builder screen opens an embedded web page hosted by Supertone where you can design custom voice references; the resulting reference can be used inside the app.


How your data is handled

The developer does not operate a backend server or analytics system for this app. The app contains no sign-in feature and no user account. Prompts you enter, voice recordings you make, and audio files the app generates are stored only in the app's private storage on your device.

The app does make outbound network requests for three specific purposes: downloading the Supertonic 3 model files from Hugging Face on first launch, displaying advertisements through Google AdMob, and loading the optional Voice Builder web page from Supertone if you open that screen.

Full details are in the privacy policy at https://ghyeok.io/voxtonic/privacy/.


Permissions

Internet — required to download the Supertonic 3 model on first launch, load ads, and load the optional Voice Builder web page.

Microphone — used only when you start a recording inside the app to create a voice reference.

Read audio files — used only when you import an existing audio file as a voice reference through the system file picker.

Wake lock — held during synthesis so the inference process is not interrupted.


Monetization

Voxtonic is free to download. It shows banner, interstitial, and native advertisements provided by Google AdMob.


Performance note

Synthesis runs on the device's CPU through ONNX Runtime. Generation time depends on the length of the input text and the performance of your device.


Trademarks

"Supertone" and "Supertonic" are trademarks of their respective owners. The Supertonic 3 model is licensed under its original Hugging Face repository terms. Voxtonic is not an official Supertone product.


Contact

Developer: GwonHyeok (GHPlanet)
Email: [email protected]
Website: https://ghyeok.io/
Privacy policy: https://ghyeok.io/voxtonic/privacy/