rysasahrialSherpaTTS is a powerful Android text‑to‑speech (TTS) engine powered by next‑generation Kaldi, designed for high‑quality offline speech. With SherpaTTS, developers and users can integrate AI voice synthesis directly into Android devices. The engine supports customizable voices—either Piper Voices or Coqui models—allowing users to tailor voice characteristics such as gender, accent, and tone.
On first launch, SherpaTTS downloads your selected voice model from Hugging Face via an internet connection. Once downloaded, all text‑to‑speech operations run entirely offline, ensuring privacy and fast execution. This offline capability makes SherpaTTS ideal for mobile environments with limited or no connectivity. End users can initiate voice tests right from the app interface to preview speech output quality.
The SherpaTTS source code is written primarily in Kotlin (≈ 76%), with supplemental Java modules (≈ 24%). It is distributed under the GNU GPL‑3.0 license, ensuring open‑source freedom. Contributions are welcome, including translations via toolate.othing.xyz and code enhancements. The codebase leverages the Sherpa‑ONNX project (Apache‑2.0) and eSpeak‑NG data (GPL‑3.0), merging technologies to support diverse voices. Voice models (Piper or Coqui) are self‑hosted on Hugging Face and downloaded on demand.
From version 1.4 upwards, the app allows easy deletion and replacement of voice models, plus the option to adjust volume, pitch, and speech speed directly via Android’s TTS system settings . Version 1.9 (released May 26 2025) includes bug fixes, UI polish, and improved translation support.
Developer Highlights:
- Integration with Android TTS API: Install SherpaTTS as a system engine and use its voices in any Android text‑to‑speech context—apps like TTS Util and accessibility services.
- Multi‑language support: Users can download separate language models (e.g., English, German) and switch seamlessly within the app. The engine enforces one active model per language but allows easy updates or replacements via app UI or beta APK.
- Privacy‑first design: No internet is needed after the first download, with all processing done on device.
SherpaTTS meets the needs of developers seeking offline, high‑quality TTS on Android. It’s ideal for mobile apps, assistive tech, language learning, and on‑device AI voice applications. The source repository includes Android implementation in Kotlin/Java, Gradle build scripts, multilingual support, Hugging Face auto‑download integration, and fastlane metadata.
- Author: woheller69 (GitHub)
- Programming Languages: Kotlin, Java
- License: GNU GPL‑3.0 (app), Apache‑2.0 (Sherpa‑ONNX core), GPL‑3.0 (eSpeak‑NG data)
- Usage Restrictions:
- Must comply with GPL‑3.0 viral licensing—any redistribution must include source and same license.
- Sherpa‑ONNX components under Apache‑2.0 enforce attribution but allow broader reuse.
- Voice model downloads (Piper or Coqui) are subject to their respective model licenses via Hugging Face.
- GitHub Repository: https://github.com/woheller69/ttsEngine