English · 한국어

Supertonic TTS

Turn any text into natural speech, entirely on your phone, and send it to whatever speaker you choose.

Supertonic TTS is an Android app that synthesizes spoken audio from text using the on-device Supertonic 3 model — no cloud, no account, no network once the voices are downloaded. Pick a voice, pick an output device, and press play — or save the result as an audio file.

When would you use this?

• You need text read aloud in someone else's voice, not your own. Paste a message, an article, or a note and let the app speak it clearly — handy for narration, accessibility, language practice, or hands-free listening.
• You need to "say" something out loud in a quiet place. In a library, a meeting, or a late-night room, type what you want to communicate and route it to a Bluetooth earphone or speaker so it's delivered as speech without you having to talk.

Features

• 🔊 On-device synthesis — runs the Supertonic 3 model locally; works fully offline after the one-time model download.
• 🎚️ Choose your output device — play through the phone speaker, a wired headset, a USB DAC, or a specific connected Bluetooth speaker/earphone.
• 💾 Save to a file — export the audio as WAV, AAC (M4A), Opus (OGG), or MP3 to anywhere you like via the system file picker.
• 🌍 Multilingual — 31 languages plus an Auto mode that handles text without a language tag.
• 🗣️ 10 preset voices — five male (M1–M5) and five female (F1–F5).
• 🎛️ Tunable — adjust quality (denoising steps) and speech speed.
• 🎨 Clean Material 3 (Jetpack Compose) interface.

Synthesize & play	Choose the output device	Export formats	Advanced settings

Requirements

• Android 9.0 (API 28) or newer.
• An arm64-v8a device (virtually all modern phones).
• ~400 MB free storage for the voice models, downloaded once on first launch.
• Internet connection for the first run only (to download the models); offline afterwards.

How to use

1. Install the app. Download supertonic-v0.1-arm64-v8a.apk from the Releases page and open it. You may need to allow installation from unknown sources.
2. Download the voices. On first launch a dialog appears and fetches the model files (~398 MB). This happens only once.
3. Type your text in the text box. Use the ✕ button to clear it.
4. Pick a voice from the dropdown.
5. Press Play ▶ to hear it, or Save 💾 to export an audio file.
6. Open “추가 설정 / Advanced settings” to choose the:
   • Output device — where the audio plays (phone, Bluetooth, wired, USB…).
   • Save format — WAV / AAC / Opus / MP3.
   • Language — a specific language or Auto.
   • Quality and Speed.

The provided APK is signed with a debug key for easy installation and testing. It is not yet intended for Play Store distribution.

Technical overview

• Model & inference. Uses the Supertonic 3 text-to-speech model via ONNX Runtime (onnxruntime-android). Synthesis is a four-model pipeline — duration predictor → text encoder → vector estimator (iterative denoising) → vocoder — producing 44.1 kHz mono float PCM. The Kotlin engine is a faithful port of the official reference implementation.
• Model delivery. The four ONNX models, voice styles, and config are downloaded from the Hugging Face repo on first run and cached in app-private storage (resumable, integrity-checked downloads).
• Playback & routing. PCM is streamed through an AudioTrack (ENCODING_PCM_FLOAT); the chosen sink is enumerated via AudioManager.getDevices() and selected with AudioTrack.setPreferredDevice().
• Audio export. Non-WAV formats are encoded with FFmpegKit; WAV is written directly. Output is saved through the Storage Access Framework.
• UI. Single-screen Jetpack Compose (Material 3) app backed by a ViewModel with Kotlin coroutines.
• Build. Kotlin 2.2 · AGP 9 · minSdk 28 / targetSdk 36. Release APKs are split per ABI and shipped as arm64-v8a only.

Credits & license

Built on the open-weight Supertonic model by Supertone Inc. The model is distributed under its own license (OpenRAIL-M) — see the model card. This app's source code is provided as-is for the Supertonic Android client.