supertonic-rs | On-Device Neural TTS for Rust

A Rust workspace providing high-quality, on-device text-to-speech via Supertonic's 99M-parameter model running on ONNX Runtime. No cloud, no API calls, no GPU required.

Crates

Crate	Description	crates.io
st-tts	High-level API — one-liner synthesize, plus core/backend/store modules for lower-level access
tauri-plugin-supertonic	Tauri v2 plugin for desktop apps

Thank you Traun Leyden for the wiki:

https://deepwiki.com/TheoSlater/supertonic-rs/1-overview

Use st-tts in a Rust project

Install:

cargo add st-tts
cargo add tokio --features macros,rt-multi-thread
cargo add anyhow

Minimal app:

use st_tts::Tts;

#[tokio::main]
async fn main() -> anyhow::Result<()> {
    // First run downloads the model into the OS data directory.
    let tts = Tts::new("Supertone/supertonic-3", "M1").await?;

    let wav = tts.synthesize_wav("Hello world", "en", None).await?;
    std::fs::write("hello.wav", wav)?;

    Ok(())
}

Need raw PCM instead of WAV:

let result = tts.synthesize("Hello world", "en", None).await?;
// result.audio: Vec<f32>
// result.sample_rate: u32

Use in a Tauri v2 project

Add the Rust plugin:

cd src-tauri
cargo add tauri-plugin-supertonic

Register it:

// src-tauri/src/lib.rs
pub fn run() {
    tauri::Builder::default()
        .plugin(tauri_plugin_supertonic::init())
        .run(tauri::generate_context!())
        .expect("error while running tauri application");
}

Allow plugin commands in your Tauri capability:

src-tauri/capabilities/default.json:

{
  "permissions": [
    "core:default",
    "supertonic:default"
  ]
}

Call it from the frontend:

npm add tauri-plugin-supertonic-api

import { loadModel, synthesize } from "tauri-plugin-supertonic-api";

await loadModel("Supertone/supertonic-3", "M1", (event) => {
  console.log(event.file, event.bytesDownloaded, event.totalBytes);
});

const { wavBase64 } = await synthesize("Hello world", "en");
new Audio(`data:audio/wav;base64,${wavBase64}`).play();

loadModel must run before synthesize. The first run downloads the model; later runs reuse the cache.

Use the OpenTUI CLI

The CLI is an interactive terminal app for typing text, choosing language, voice, and speed, then either hearing the generated speech or saving it as a WAV file.

Requirements:

curl -fsSL https://bun.sh/install | bash
rustup toolchain install stable

For playback inside the app, install one of these audio players: pw-play, paplay, aplay, or ffplay.

Run:

cd supertonic-cli
bun install
bun run src/index.tsx

Controls:

Tab          move between fields
Enter        open synthesize popup
Tab/arrows   switch between Hear in app and Save WAV file
Enter        confirm popup choice
Esc          cancel popup or quit
Ctrl+C       quit

The first synth downloads Supertone/supertonic-3. Save mode writes supertonic-cli/supertonic.wav; hear mode writes a temporary preview and plays it.

Publishing Order

tauri-plugin-supertonic depends on st-tts. Publish in this order:

cargo publish -p st-tts
cargo publish -p tauri-plugin-supertonic