VoxSherpa TTS is an open-source Android app that runs studio-quality neural text-to-speech entirely on-device, with no cloud calls, accounts, or usage limits. It is listed in the official README of k2-fsa/sherpa-onnx, the on-device inference engine it is built on, and offers two neural engines: Kokoro-82M for higher-quality, multilingual speech (Hindi, English, British English, French, Spanish, Chinese, Japanese, and 50+ languages) and Piper/VITS for fast, lightweight synthesis on lower-end hardware. Version 4.0 adds multi-speaker dialogue synthesis, letting a single script assign different voices to different lines via a [speaker:N] tag system with adjustable tone and pacing, aimed at audiobooks, podcasts, and narrated dialogue. Additional features include PDF/TXT-to-audio conversion, system-wide TTS integration for any Android app, custom ONNX model import, and local generation history with export to WAV.