New Features - One-click copy API address: Tap to copy the intranet pronunciation service address for easy pasting into external apps like Legado Reader - Custom HTTP API port: Supports custom port numbers (1024–65535), no longer hardcoded to 8080 to avoid conflicts with other apps - Newly created/imported voices automatically appear at the top - Audio mixer menu consolidation: The ⋮ menu on the voice page has been reduced from 5 to 3 items (Add Voice / Import-Export / Batch Management) for clearer operation - Batch modify inference steps: Supports selecting multiple voices to unify the num_steps setting - Drag-and-drop voice reordering: Long press on the voice name area to drag and reorder, with scaling + shadow feedback while dragging - Inference steps tooltip: Added num_steps explanation when creating or editing a voice - Auto-scrolling log console: Automatically scrolls to the bottom when new logs arrive - Improved system language compatibility: Added recognition for the cmn (Mandarin) ISO 639-3 language code, fixing the issue where the app could not be set as the system TTS on some devices Core Engine Upgrade - ~4.4% inference performance improvement - Completely fixed crashes on Snapdragon 8 Elite: ORT 1.24.4 includes the KleidiAI SME2 fix, no longer triggering illegal instruction crashes - Full-architecture APK (4 ABIs): Restored packaging for arm64-v8a, armeabi-v7a, x86, and x86_64, covering the vast majority of physical devices and emulators Performance Optimizations - Instant speed adjustment across three paths: Speed changes for in-app preview, system TTS, and HTTP API all take effect immediately (<10ms), no longer requiring re-synthesis - Reduced HTTP API speed change latency: Takes <10ms Bug Fixes - Fixed emulator crashes: No longer crashes on startup in x86/x86_64 emulators - Fixed percent sign reading: "95%" is now correctly read as "ninety-five percent" - Fixed HTTP API sentence skipping: Sentences containing % are no longer skipped and silent (requires re-importing Legado configuration) - Fixed HTTP API speed cap: Now properly clamped to the 2.0x upper limit - Improved short sentence pronunciation: Very short sentences of 2–4 characters are no longer muffled or distorted - Fixed system TTS first sentence delay: The issue where the first sentence of long paragraphs was generated too slowly has been resolved - Intelligent sentence splitting and merging: Overly short fragments are automatically merged (≥10 characters), while overly long fragments are split at punctuation (≤50 characters) - Fixed number reading: Abbreviated years like "23年8月" are correctly read as "August twenty-third", and leading zeros like "09年5月" are read as "May zero-nine"

CloneTTS is an offline, native Text-to-Speech (TTS) engine for Android devices. It allows you to clone desired voices and use them directly to read books or long texts on your phone — no internet connection required. All inference runs locally on-device. Key Features - Offline Voice Cloning: Record 1–3 seconds of speech to create a custom voice. - System TTS: Registers as an Android system TTS engine, compatible with Legado, Moon+ Reader, etc. - HTTP API Server: Built-in local HTTP server supporting GET and POST requests. - Speed & Volume Control: 0.5x–2.0x speech rate, 0%–200% volume. - Pronunciation Correction: Add powerful custom plain text or regex-based substitution rules. - Voice Backup: ZIP-based export/import with append or overwrite modes. Supported Languages · Chinese ℹ️ Info: No additional languages are on the roadmap for the short term.