Supertonic

https://github.com/supertone-oss-archive/supertonic
Admin
Supertonic is a lightning-fast, on-device multilingual text-to-speech system designed for local...
Tokens:25,989
Snippets:246
Trust Score:4.4
License:MIT
Update:1 month ago
Show doc for...
Tokens:
Raw
### Run browser example

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/README.md

Run the browser example. Requires the assets/ directory from the repository root. Installs dependencies and starts the development server.

```bash
cd web
npm install
npm run dev
```

--------------------------------

### Run Node.js example

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/README.md

Run the Node.js example. Requires the assets/ directory from the repository root. Installs dependencies and starts the application.

```bash
cd nodejs
npm install
npm start
```

--------------------------------

### Run Go example

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/README.md

Run the Go example. Requires the ONNX Runtime C library (e.g., brew install onnxruntime on macOS) and the assets/ directory from the repository root.

```bash
cd go
go mod download
go run example_onnx.go helper.go
```

--------------------------------

### Install preserved Python SDK and run example_pypi.py

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/README.md

Installs the preserved Python SDK source with Supertonic 3 support and runs the SDK example. The example uses the local assets/ directory and does not download models.

```bash
python -m pip install "git+https://github.com/supertone-oss-archive/supertonic-py.git@df0f9686dac7fbbde391b759e2ee5286a3737622"
python py/example_pypi.py
```

--------------------------------

### Install dependencies and run ONNX speech generation example

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/README.md

Installs Python dependencies and runs the ONNX example to generate speech locally. The output WAV is saved in py/results/; after dependencies and models are downloaded, inference works without a network connection.

```bash
python -m pip install -r py/requirements.txt
cd py
python example_onnx.py --n-test 1 --text "This speech was generated locally with the archived Supertonic model." --lang en
```

--------------------------------

### Build the C++ example

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/cpp/README.md

Build the example by running cmake and building the Release configuration, then run the executable.

```bash
cd cpp
mkdir -p build && cd build
cmake ..
cmake --build . --config Release
./example_onnx

```

--------------------------------

### Start development server

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/web/README.md

Run this command to start the local development server (usually at http://localhost:3000).

```bash
npm run dev
```

--------------------------------

### Install dependencies with dotnet restore

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/csharp/README.md

Restores the project dependencies. Required before running the example.

```bash
dotnet restore
```

--------------------------------

### Run Swift example

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/README.md

Run the Swift example. Requires the assets/ directory from the repository root. Builds in release mode and runs the executable.

```bash
cd swift
swift build -c release
.build/release/example_onnx
```

--------------------------------

### Run C# example

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/README.md

Run the C# example. Targets .NET 9 with major-version roll-forward allowed. Requires the assets/ directory from the repository root.

```bash
cd csharp
dotnet restore
dotnet run
```

--------------------------------

### Install Rust

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/rust/README.md

Use this command to install Rust if it is not already present on your system.

```bash
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
```

--------------------------------

### Run Java example

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/README.md

Run the Java example. Requires a JDK (not just a JRE) and the assets/ directory from the repository root. Builds and executes the application.

```bash
cd java
mvn clean install
mvn exec:java
```

--------------------------------

### Install dependencies with uv sync

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/py/README.md

Installs project dependencies using uv.

```bash
uv sync
```

--------------------------------

### Install ONNX Runtime on macOS with Homebrew

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/go/README.md

Installs the ONNX Runtime C library on macOS via Homebrew. The Go example auto-detects Homebrew ONNX Runtime paths on both Apple Silicon and Intel Macs.

```bash
brew install onnxruntime
```

--------------------------------

### Install dependencies with Maven

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/java/README.md

Runs Maven clean install to build the project and install dependencies.

```bash
mvn clean install
```

--------------------------------

### Install dependencies with pip

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/py/README.md

Alternative installation using pip with requirements.txt.

```bash
pip install -r requirements.txt
```

--------------------------------

### Install dependencies (Windows vcpkg)

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/cpp/README.md

Install nlohmann-json and onnxruntime for x64-windows using vcpkg, then integrate with the build system.

```powershell
vcpkg install nlohmann-json:x64-windows onnxruntime:x64-windows
vcpkg integrate install

```

--------------------------------

### Run Rust example

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/README.md

Run the Rust example. Requires the assets/ directory from the repository root. Builds in release mode and runs the executable.

```bash
cd rust
cargo build --release
./target/release/example_onnx
```

--------------------------------

### Run C++ example

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/README.md

Run the C++ example. Requires the assets/ directory from the repository root. Builds with CMake and runs the executable.

```bash
cd cpp
mkdir build && cd build
cmake .. && cmake --build . --config Release
./example_onnx
```

--------------------------------

### Install dependencies

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/web/README.md

Run this command to install the required npm dependencies for the project.

```bash
npm install
```

--------------------------------

### Install uv

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/py/README.md

Installs the uv package manager using the official install script.

```bash
curl -LsSf https://astral.sh/uv/install.sh | sh
```

--------------------------------

### Run iOS example

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/README.md

Run the iOS example. Requires the assets/ directory from the repository root. Generates the Xcode project and opens it. After opening, select your Team and iPhone as run destination in Xcode.

```bash
cd ios/ExampleiOSApp
xcodegen generate
open ExampleiOSApp.xcodeproj
```

--------------------------------

### Build and run the TTS example binary

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/go/README.md

Builds the standalone executable from the Go source files, then runs it with a voice style and text argument. The build command must be run first to produce the binary.

```bash
go build -o tts_example example_onnx.go helper.go
```

```bash
./tts_example -voice-style "../assets/voice_styles/M1.json" -text "Hello world"
```

--------------------------------

### Clone repository and set up Python 3.11 environment

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/README.md

Sets up a Python 3.11 virtual environment and installs the Hugging Face Hub client. Run these commands from the repository root before downloading the archived model.

```bash
git clone https://github.com/supertone-oss-archive/supertonic.git
cd supertonic
python3.11 -m venv .venv
source .venv/bin/activate
python -m pip install huggingface_hub
```

--------------------------------

### Install dependencies (macOS)

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/cpp/README.md

Install cmake, nlohmann-json, and onnxruntime using Homebrew.

```bash
brew install cmake nlohmann-json onnxruntime

```

--------------------------------

### Run example with default settings

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/cpp/README.md

Runs inference using default voice style, text, output directory, total steps (8), and number of generations (4).

```bash
./example_onnx

```

--------------------------------

### Prepare Assets for iOS Target

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/ios/README.md

Create the necessary directories and sync the ONNX models and voice styles from the repository root into the iOS example app directory.

```bash
cd ios/ExampleiOSApp
mkdir -p onnx voice_styles
rsync -a ../../assets/onnx/ onnx/
rsync -a ../../assets/voice_styles/ voice_styles/
```

--------------------------------

### Install dependencies and ONNX Runtime (Ubuntu/Debian)

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/cpp/README.md

Install CMake, g++, nlohmann-json, and download/extract ONNX Runtime v1.16.3 (linux-x64), then copy headers and libs to system directories and run ldconfig.

```bash
sudo apt-get install -y cmake g++ nlohmann-json3-dev
wget https://github.com/microsoft/onnxruntime/releases/download/v1.16.3/onnxruntime-linux-x64-1.16.3.tgz
tar -xzf onnxruntime-linux-x64-1.16.3.tgz
sudo cp -r onnxruntime-linux-x64-1.16.3/include/* /usr/local/include/
sudo cp -r onnxruntime-linux-x64-1.16.3/lib/* /usr/local/lib/
sudo ldconfig

```

--------------------------------

### Install Node.js dependencies

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/nodejs/README.md

Install the Node.js package dependencies. Requires Node.js v16 or higher and npm or yarn.

```bash
cd nodejs
npm install
```

--------------------------------

### Install ONNX Runtime on Linux from GitHub releases

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/go/README.md

Downloads and installs ONNX Runtime v1.16.0 on Linux. Copies libraries and headers to /usr/local and runs ldconfig to update the linker cache.

```bash
# Download ONNX Runtime from GitHub releases
wget https://github.com/microsoft/onnxruntime/releases/download/v1.16.0/onnxruntime-linux-x64-1.16.0.tgz
tar -xzf onnxruntime-linux-x64-1.16.0.tgz
sudo cp onnxruntime-linux-x64-1.16.0/lib/* /usr/local/lib/
sudo cp -r onnxruntime-linux-x64-1.16.0/include/* /usr/local/include/
sudo ldconfig
```

--------------------------------

### Install XcodeGen via Homebrew

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/ios/README.md

Install the required XcodeGen tool using Homebrew if it is not already present on your system.

```bash
brew install xcodegen
```

--------------------------------

### Install OpenJDK 17 and Maven on macOS with Homebrew

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/java/README.md

Installs OpenJDK 17 and Maven via Homebrew and sets JAVA_HOME and PATH environment variables for the current shell session.

```bash
brew install openjdk@17 maven
export JAVA_HOME="$(brew --prefix openjdk@17)/libexec/openjdk.jdk/Contents/Home"
export PATH="$(brew --prefix openjdk@17)/bin:$PATH"
```

--------------------------------

### Run default inference with Maven exec plugin

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/java/README.md

Executes the TTS example with default settings, using voice style M1.json, a default text, output directory results/, 8 total steps, and 4 generations.

```bash
mvn exec:java
```

--------------------------------

### Run TTS JAR directly with apostrophe escaping

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/java/README.md

Runs the TTS example JAR directly to avoid shell escaping issues with apostrophes in the text.

```bash
java -jar target/tts-example.jar --total-step 10 --text "Text with apostrophe's here"
```

--------------------------------

### Long-form inference example

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/cpp/README.md

Runs with a long text; the system automatically chunks (max 300 chars) and combines into a single audio file, with 0.3s silences between chunks. Note: --batch disables chunking.

```bash
./example_onnx \
  --voice-style ../../assets/voice_styles/M1.json \
  --text "Once upon a time, in a small village nestled between rolling hills, there lived a young artist named Eleanor. Every morning, she would wake up before dawn to capture the first light of day. The golden rays streaming through her window inspired countless paintings. Her work was known throughout the region for its rich colors and emotional depth. People from far and wide came to see her gallery, and many said her paintings could tell stories that words could never."

```

--------------------------------

### Batch inference example

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/cpp/README.md

Uses --batch to process multiple voice-text pairs: first pair with M1.json and English text, second with F1.json and Korean text. Automatic text chunking is disabled in batch mode.

```bash
./example_onnx \
  --voice-style ../../assets/voice_styles/M1.json,../../assets/voice_styles/F1.json \
  --text "The sun sets behind the mountains, painting the sky in shades of pink and orange.|오늘 아침에 공원을 산책했는데, 새소리와 바람 소리가 너무 좋아서 한참을 멈춰 서서 들었어요." \
  --lang en,ko \
  --batch

```

--------------------------------

### Run high quality inference with example_onnx.js

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/nodejs/README.md

Run inference with 10 denoising steps instead of the default 8 for higher quality output, at the cost of slower inference.

```bash
node example_onnx.js \
  --total-step 10 \
  --voice-style "../assets/voice_styles/M1.json" \
  --text "Increasing the number of denoising steps improves the output's fidelity and overall quality."
```

--------------------------------

### Build the Project

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/rust/README.md

Compile the project in release mode using Cargo.

```bash
cargo build --release
```

--------------------------------

### Build and run Supertonic in Release mode

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/csharp/README.md

Build the project in Release configuration, then run the compiled executable. The executable path is relative to the project root after a successful build.

```bash
dotnet build -c Release
```

```bash
./bin/Release/net9.0/Supertonic
```

--------------------------------

### Run the Flutter demo on macOS

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/flutter/README.md

Run these commands from the repository root to clean, fetch dependencies, and launch the Flutter demo on macOS. Requires Flutter SDK version ^3.5.0.

```bash
flutter clean
flutter pub get
flutter run -d macos
```

--------------------------------

### Run default inference with example_onnx

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/swift/README.md

Runs inference with default settings: voice style M1.json, a fixed English text, output directory results/, 8 total steps, and 4 generations.

```bash
.build/release/example_onnx
```

--------------------------------

### Generate and Open Xcode Project

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/ios/README.md

Use XcodeGen to generate the project file and open it in Xcode.

```bash
xcodegen generate
open ExampleiOSApp.xcodeproj
```

--------------------------------

### Build and Run Fat JAR

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/java/README.md

Commands to build a fat JAR with Maven and run it with or without arguments.

```bash
mvn clean package
```

```bash
java -jar target/tts-example.jar
```

```bash
java -jar target/tts-example.jar --total-step 10 --text "Your custom text here"
```

--------------------------------

### Download Go module dependencies

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/go/README.md

Downloads the Go module dependencies required by the project.

```bash
go mod download
```

--------------------------------

### Run default inference with dotnet run

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/csharp/README.md

Runs inference with default settings: voice style M1.json, a fixed English text, output to results/, 8 total steps, and 4 generations.

```bash
dotnet run
```

--------------------------------

### Run high quality inference with Maven exec plugin

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/java/README.md

Increases denoising steps to 10 for higher quality output, at the cost of slower inference.

```bash
mvn exec:java -Dexec.args="--total-step 10 --voice-style ../assets/voice_styles/M1.json --text 'Increasing the number of denoising steps improves the output fidelity and overall quality.'"
```

--------------------------------

### Run batch inference with example_onnx.js

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/nodejs/README.md

Run batch inference with multiple voice styles and texts. The --batch flag enables batch processing, generates speech for 2 voice-text pairs, and disables automatic text chunking.

```bash
node example_onnx.js \
  --voice-style "../assets/voice_styles/M1.json,../assets/voice_styles/F1.json" \
  --text "The sun sets behind the mountains, painting the sky in shades of pink and orange.|오늘 아침에 공원을 산책했는데, 새소리와 바람 소리가 너무 좋아서 한참을 멈춰 서서 들었어요." \
  --lang "en,ko" \
  --batch
```

--------------------------------

### Run high-quality inference with 10 denoising steps

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/go/README.md

Runs high-quality inference with 10 denoising steps instead of the default 8. Produces higher quality output at the cost of slower inference.

```bash
go run example_onnx.go helper.go \
  -total-step 10 \
  -voice-style "../assets/voice_styles/M1.json" \
  -text "Increasing the number of denoising steps improves the output's fidelity and overall quality."
```

--------------------------------

### Run high-quality inference with --total-step 10

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/swift/README.md

Increases denoising steps to 10 instead of the default 8 for higher quality output, at the cost of slower inference.

```bash
.build/release/example_onnx \
  --total-step 10 \
  --voice-style ../assets/voice_styles/M1.json \
  --text "Increasing the number of denoising steps improves the output's fidelity and overall quality."
```

--------------------------------

### Build the project with swift build -c release

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/swift/README.md

Builds the project in release mode using Swift Package Manager. Requires Swift 5.9 or later and macOS 13.0 or later.

```bash
swift build -c release
```

--------------------------------

### Run High Quality Inference

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/rust/README.md

Increase the denoising steps to improve output fidelity at the cost of slower inference speed.

```bash
# Using cargo run
cargo run --release --bin example_onnx -- \
  --total-step 10 \
  --voice-style ../assets/voice_styles/M1.json \
  --text "Increasing the number of denoising steps improves the output's fidelity and overall quality."

# Or using the binary directly
./target/release/example_onnx \
  --total-step 10 \
  --voice-style ../assets/voice_styles/M1.json \
  --text "Increasing the number of denoising steps improves the output's fidelity and overall quality."
```

--------------------------------

### Run batch inference with Maven exec plugin

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/java/README.md

Processes multiple voice styles and texts in a single batch. Uses M1.json for English text and F1.json for Korean text, with language codes en and ko.

```bash
mvn exec:java -Dexec.args="--batch --voice-style ../assets/voice_styles/M1.json,../assets/voice_styles/F1.json --text 'The sun sets behind the mountains, painting the sky in shades of pink and orange.|오늘 아침에 공원을 산책했는데, 새소리와 바람 소리가 너무 기분 좋았어요.' --lang en,ko"
```

--------------------------------

### Download archived Supertonic 3 model with hf download

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/README.md

Downloads the archived Supertonic 3 model files (ONNX models, configuration, and preset voice styles) into the assets/ directory. The revision pins the archived snapshot; Git LFS is not required.

```bash
hf download supertone-oss-archive/supertonic-3 \
  --revision aafc6e32416a594460b32413efc49d7fe4ce6d46 \
  --local-dir assets
```

--------------------------------

### Cite SupertonicTTS main architecture (arXiv:2503.23108)

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/README.md

BibTeX entry for the SupertonicTTS main architecture paper. Use when citing the overall system in academic work.

```bibtex
@article{kim2025supertonic,
  title={SupertonicTTS: Towards Highly Efficient and Streamlined Text-to-Speech System},
  author={Kim, Hyeongju and Yang, Jinhyeok and Yu, Yechan and Ji, Seunghun and Morton, Jacob and Bous, Frederik and Byun, Joon and Lee, Juheon},
  journal={arXiv preprint arXiv:2503.23108},
  year={2025},
  url={https://arxiv.org/abs/2503.23108}
}
```

--------------------------------

### Run Default Inference

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/rust/README.md

Execute inference with default settings using either cargo run or the compiled binary.

```bash
# Using cargo run
cargo run --release --bin example_onnx

# Or directly execute the built binary (faster)
./target/release/example_onnx
```

--------------------------------

### Run long-form inference with example_onnx.js

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/nodejs/README.md

Run inference on a long text. The system automatically chunks the text into segments (max 300 characters by default), inserts 0.3-second silences between chunks, and combines them into a single audio file. Note: automatic chunking is disabled in batch mode.

```bash
node example_onnx.js \
  --voice-style "../assets/voice_styles/M1.json" \
  --text "Once upon a time, in a small village nestled between rolling hills, there lived a young artist named Clara. Every morning, she would wake up before dawn to capture the first light of day. The golden rays streaming through her window inspired countless paintings. Her work was known throughout the region for its vibrant colors and emotional depth. People from far and wide came to see her gallery, and many said her paintings could tell stories that words never could."
```

--------------------------------

### Manually set ONNXRUNTIME_LIB_PATH for Linux or macOS

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/go/README.md

Manually sets the ONNXRUNTIME_LIB_PATH environment variable to a custom library path for Linux or macOS.

```bash
export ONNXRUNTIME_LIB_PATH=/path/to/libonnxruntime.so  # Linux
# or
export ONNXRUNTIME_LIB_PATH=/path/to/libonnxruntime.dylib  # macOS
```

--------------------------------

### Run batch inference with --batch flag

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/swift/README.md

Processes multiple voice styles and texts in one batch. Uses M1.json for English and F1.json for Korean, with language tags en and ko. Generates speech for 2 voice-text-language triplets.

```bash
.build/release/example_onnx \
  --batch \
  --voice-style ../assets/voice_styles/M1.json,../assets/voice_styles/F1.json \
  --text "The sun sets behind the mountains, painting the sky in shades of pink and orange.|오늘 아침에 공원을 산책했는데, 새소리와 바람 소리가 너무 기분 좋았어요." \
  --lang en,ko
```

--------------------------------

### Run Batch Inference

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/rust/README.md

Process multiple voice styles and texts simultaneously by providing comma-separated values.

```bash
# Using cargo run
cargo run --release --bin example_onnx -- \
  --batch \
  --voice-style ../assets/voice_styles/M1.json,../assets/voice_styles/F1.json \
  --text "The sun sets behind the mountains, painting the sky in shades of pink and orange.|오늘 아침에 공원을 산책했는데, 새소리와 바람 소리가 너무 기분 좋았어요." \
  --lang en,ko

# Or using the binary directly
./target/release/example_onnx \
  --batch \
  --voice-style ../assets/voice_styles/M1.json,../assets/voice_styles/F1.json \
  --text "The sun sets behind the mountains, painting the sky in shades of pink and orange.|오늘 아침에 공원을 산책했는데, 새소리와 바람 소리가 너무 기분 좋았어요." \
  --lang en,ko
```

--------------------------------

### Set ONNXRUNTIME_LIB_PATH automatically for macOS and Linux

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/go/README.md

Automatically detects and sets the ONNXRUNTIME_LIB_PATH environment variable for macOS and Linux. Recommended when the library is not in a standard or Homebrew location.

```bash
# macOS
export ONNXRUNTIME_LIB_PATH=$(brew --prefix onnxruntime 2>/dev/null)/lib/libonnxruntime.dylib

# Linux
export ONNXRUNTIME_LIB_PATH=$(find /usr/local/lib /usr/lib -name "libonnxruntime.so*" 2>/dev/null | head -n 1)
```

--------------------------------

### Cite Length-Aware RoPE paper (arXiv:2509.11084)

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/README.md

BibTeX entry for the Length-Aware RoPE (LARoPE) paper. Use when citing the text-speech alignment technique.

```bibtex
@article{kim2025larope,
  title={Length-Aware Rotary Position Embedding for Text-Speech Alignment},
  author={Kim, Hyeongju and Lee, Juheon and Yang, Jinhyeok and Morton, Jacob},
  journal={arXiv preprint arXiv:2509.11084},
  year={2025},
  url={https://arxiv.org/abs/2509.11084}
}
```

--------------------------------

### Run long-form inference with Maven exec plugin

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/java/README.md

Automatically chunks long text, synthesizes each chunk, and concatenates with 0.3-second pauses. This is the default behavior when not using --batch.

```bash
mvn exec:java -Dexec.args="--voice-style ../assets/voice_styles/M1.json --text 'This is a very long text that will be automatically split into multiple chunks. The system will process each chunk separately and then concatenate them together with natural pauses between segments. This ensures that even very long texts can be processed efficiently while maintaining natural speech flow and avoiding memory issues.'"
```

--------------------------------

### Run batch inference with --batch flag

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/csharp/README.md

Runs batch inference for two voice-text pairs: M1.json with English text and F1.json with Korean text. The --batch flag enables batch mode and disables automatic text chunking.

```bash
dotnet run -- \
  --voice-style ../assets/voice_styles/M1.json,../assets/voice_styles/F1.json \
  --text "The sun sets behind the mountains, painting the sky in shades of pink and orange.|오늘 아침에 공원을 산책했는데, 새소리와 바람 소리가 너무 좋아서 한참을 멈춰 서서 들었어요." \
  --lang en,ko \
  --batch
```

--------------------------------

### Cite RobustSpeechFlow paper (arXiv:2605.22083)

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/README.md

BibTeX entry for the RobustSpeechFlow paper. Use when citing the augmentation-based contrastive flow matching technique.

```bibtex
@misc{yang2026robustspeechflowlearningrobusttexttospeech,
      title={RobustSpeechFlow: Learning Robust Text-to-Speech Trajectories via Augmentation-based Contrastive Flow Matching}, 
      author={Jinhyeok Yang and Hyeongju Kim and Yechan Yu and Joon Byun and Frederik Bous and Juheon Lee},
      year={2026},
      eprint={2605.22083},
      archivePrefix={arXiv},
      primaryClass={cs.SD},
      url={https://arxiv.org/abs/2605.22083}, 
}
```

--------------------------------

### Run long-form inference with automatic text chunking

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/swift/README.md

Automatically splits long texts into chunks, synthesizes each separately, and concatenates with 0.3-second pauses. This happens by default when not using --batch. Note: chunking is disabled in batch mode.

```bash
.build/release/example_onnx \
  --voice-style ../assets/voice_styles/M1.json \
  --text "This is a very long text that will be automatically split into multiple chunks. The system will process each chunk separately and then concatenate them together with natural pauses between segments. This ensures that even very long texts can be processed efficiently while maintaining natural speech flow and avoiding memory issues."
```

--------------------------------

### Cite Self-Purifying Flow Matching paper (arXiv:2509.19091)

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/README.md

BibTeX entry for the Self-Purifying Flow Matching paper. Use when citing the noisy-label training technique.

```bibtex
@article{kim2025spfm,
  title={Training Flow Matching Models with Reliable Labels via Self-Purification},
  author={Kim, Hyeongju and Yu, Yechan and Yi, June Young and Lee, Juheon},
  journal={arXiv preprint arXiv:2509.19091},
  year={2025},
  url={https://arxiv.org/abs/2509.19091}
}
```

--------------------------------

### Adjust speech speed

Source: https://github.com/supertone-oss-archive/supertonic/blob/main/py/README.md

Controls speech speed with --speed. Values >1.0 produce faster speech, <1.0 slower. Default is 1.05; recommended range is 0.9 to 1.5.

```bash
# Faster speech (speed > 1.0)
uv run example_onnx.py \
  --voice-style ../assets/voice_styles/F2.json \
  --text "This text will be synthesized at a faster pace." \
  --speed 1.2

# Slower speech (speed < 1.0)
uv run example_onnx.py \
  --voice-style ../assets/voice_styles/M2.json \
  --text "This text will be synthesized at a slower, more deliberate pace." \
  --speed 0.9
```
Tokens:
Raw