Xime - Wubi / Pinyin Input Method for Android

简体中文 · 繁體中文

Windows Version | Linux Version | Predictive Text Model | Handwriting Model

An Android input method built on the Rime engine, designed for efficient Chinese text input with Wubi (五笔) and Pinyin support.

---

This input method supports both Wubi (五笔) and Pinyin input. The author primarily uses Wubi with Pinyin as a fallback, so resources lean toward Wubi.

Full Keyboard (Light)	Full Keyboard (Dark)	Radical Swipe	Quick Actions
Floating Keyboard	T9 Pinyin	Numpad	Symbols
Handwriting	Handwriting (Candidates)	Voice Input	Emoji Keyboard
Theme Settings (Light)	Theme Settings (Dark)	Plugin Manager	Extension Store

• Multiple Input Schemas - Built-in Wubi 86/98, Pinyin, and mixed schemas; supports custom schemas (Shuangpin, Stroke, etc.) via the schema marketplace or wireless import
• Rime Engine - Powered by the mature and reliable Rime input method engine for accurate Chinese input
• Rich Keyboard Layouts - QWERTY full keyboard, T9 Pinyin, Stroke 9-key, Handwriting, Numpad (with calculator)
• Floating Keyboard - Floating card style with drag support, semi-transparent rounded design
• Voice-to-Text - Local offline ASR (built-in streaming zipformer2 engine) plus online ASR plugins (FunAsr, Volc, etc.)
• AI Enhancement - Transformer-based predictive text for faster input
• Clean UI - Material Design 3, light/dark themes with multiple color schemes
• Keyboard Adjustment - Adjustable keyboard height and position
• Toolbar Customization - Customizable toolbar button layout and functions
• Haptic Feedback - Adjustable sound and vibration intensity
• Swipe Gestures - Cursor movement, deletion, symbol input via swipe gestures
• Clipboard Manager - Clipboard history with quick send and pinning
• Clipboard Sync - Bidirectional clipboard sync with remote devices via plugins (WebDAV, ximed, etc.)
• Candidate Coding Hints - Display Wubi codes for candidates to aid learning
• Radical Display - Swipe down on keys to show Wubi radicals for memory aid
• Physical Keyboard Support - Floating candidate bar when using hardware/bluetooth keyboards
• WebDAV Sync - Backup and restore schemas and settings via WebDAV
• Plugin Marketplace - Extensible Lua plugins (emoji, clipboard sync, online ASR, etc.) via the built-in marketplace

Requirements

• Android 9.0 (API 28) or later

Installation

Download

Choose the APK matching your device architecture:

• arm64-v8a: Modern phones (recommended for most users)
• armeabi-v7a: Older 32-bit phones
• x86_64: Emulators
• universal: All architectures (larger file size)

From Releases

1. Download the latest APK from Releases
2. Install the application
3. Enable Xime in system input method settings
4. Set Xime as the current input method

Plugins (Optional)

Plugins are Lua-script plugins (.xipk format), installable and enabled from the app's Settings > Extension Store:

• kaomoji: Kaomoji text emoticons
• meme-bunny: Funny bunny sticker pack (8 stickers)
• xime-fluent-emoji: Fluent UI 3D-style emoji plugin (222 curated 3D emojis, 9 categories)
• funasr-asr: Alibaba Bailian FunAsr online speech recognition
• volc-asr: Volcano Engine online speech recognition
• webdav-clipboard-sync: WebDAV-based clipboard sync
• ximed-clipboard-sync: ximed-service-based clipboard sync

For the full plugin list, see the Plugin Center, or browse and install directly from the app's Settings > Extension Store.

Build from Source

1. Clone the project and build the APK
2. Install the application
3. Enable Xime in system input method settings
4. Set Xime as the current input method

Documentation

For detailed documentation, visit https://ime.ximei.me.

Building

# Clone with submodules
git clone --recursive https://github.com/ximeiorg/Xime.git

# Or initialize submodules in an existing clone
git submodule update --init --recursive

# Build Release APK
./gradlew assembleRelease

AI Model Download

Predictive Text Model

• Repository: https://github.com/ximeiorg/predictive-text
• Model: https://www.modelscope.cn/models/bikeand/predictive-text-small
• File: model_int8_dynamic.onnx (~17MB)
• Vocabulary: vocab.json
• Location: filesDir/ (app private directory root)
• Function: Transformer-based Chinese word prediction for intelligent candidate suggestions

Speech Recognition Model

• Model: https://www.modelscope.cn/models/bikeand/asr
• File: sherpa-onnx-streaming-zipformer-zh-int8-2025-06-30.tar.bz2 (~132MB)
• Function: Streaming zipformer2 Chinese speech recognition (offline, on-device)

Note: All models can be downloaded directly from within the app (Settings > Smart Prediction / Speech Recognition) — no manual placement required.

Tech Stack

• Kotlin
• Jetpack Compose
• Material Design 3
• Rime (librime)
• JNI (Native C++)

Contributing

Contributions welcome! Please read CONTRIBUTING.md before submitting a PR.

Core rules:

• File an Issue first — All changes require a prior discussion via an Issue
• Minimal changes — PRs must contain only the minimum changes needed
• GPG signing — All commits must be GPG-signed

Acknowledgments

• Rime - Input method engine
• Trime - Configuration reference
• fcitx5-android - Keyboard layout reference
• onnxruntime - ONNX inference runtime for predictive text and speech recognition models

License

GPLv3 License