A privacy-focused, offline-first Android EPUB reader built with Kotlin and the Readium 3.1.0 toolkit. Features three text-to-speech engines (System TTS, Piper, and Kokoro), sentence-level highlighting, bookmarks, full-text search, and a clean Material 3 UI. Features Library - Import EPUB files from device storage - Grid or list view with cover art - Drag-to-reorder via handle - Background metadata extraction (title, author, cover) - Long-press to delete Reader - Full EPUB rendering via Readium (WebView-based) - Table of Contents navigation - Reading position auto-saved and restored - Full-text search (up to 200 results) - Bookmarks: add, view, navigate, delete - Display settings: font family, font size, theme (Light/Sepia/Dark) Text-to-Speech System Quality: varies Speed: real-time Offline: no Voices: device voices Piper Quality: good Speed: fast, about 0.5x RTF Offline: yes Voices: 7 English voices Kokoro Quality: near-human Speed: slower, about 3 to 7x RTF Offline: yes Common TTS features: - Sentence-block highlighting (current + next sentence preview) - Auto-scroll and auto-page-turn - Automatic chapter advancement - Speed control (0.5x-3.0x) - Foreground service with media notification controls - Error recovery: resume from where it stopped Piper TTS (Offline): - 7 English voices (US + GB, male + female) - ~63MB per voice, download in-app - Default engine with auto-download on first play - DC offset removal and fade-in/out for clean audio Kokoro TTS (Offline, Near-Human Quality): - 3 downloadable model packs: - Kokoro English v0.19 (~340MB) -- 11 English voices - Kokoro Multi-lang v1.0 (~333MB) -- 53 voices (English + Chinese + others) - Kokoro Multi-lang v1.1 (~348MB) -- 103 voices (English + Chinese) - Per-pack download/delete with speaker selection - 24kHz sample rate output - Streaming playback via generateWithCallback -- audio starts within ~200ms - Paragraph-based generation (sentences grouped into ~400 char chunks) for seamless transitions - Parallel pre-generation of next paragraph while current one plays - Adaptive thread count (up to 8 threads on capable devices) License This project uses the following open-source components: - Readium Kotlin Toolkit (BSD-3-Clause) - sherpa-onnx (Apache-2.0) - Piper TTS models (MIT) - Kokoro TTS (Apache-2.0)