Overview
Key Features
Professional Speech Transcription
Real-time and batch transcription modes with multi-engine architecture covering Chinese, English, Japanese, Korean and more. Auto punctuation, emotion detection, and language identification.
On-Device AI Assistant
Powered by Gemma 4 on-device LLM with no cloud API and zero privacy risk. Auto-generates summaries and to-do items, supports voice and text conversation.
Privacy-First Design
All speech recognition and AI inference run entirely on-device. No account registration, no login, no tracking. Uninstalling deletes all data automatically.
Blazing Fast Experience
whisper.cpp NEON-optimized, SenseVoice single transcription in ~2 seconds. Background transcription with notification progress bar.
Multilingual Support
UI in Simplified Chinese and English. Transcription supports Chinese, English, Japanese, Korean, plus Cantonese and Wu Chinese dialects.
Smart File Management
Unified file management for recordings and transcripts. Cached AI summaries open instantly. Share via system share sheet and re-transcribe with different models.
Technical Highlights
| STT Engine 1 | SenseVoice (zh/en/ja/ko/yue) — 94%+ Chinese accuracy |
| STT Engine 2 | whisper.cpp (multilingual, NEON-optimized) |
| STT Engine 3 | Paraformer (Chinese-optimized) |
| STT Engine 4 | FunASR Nano (Chinese dialects) |
| LLM Engine | Gemma 4 E2B/E4B via LiteRT-LM (on-device) |
| Acceleration | GPU (OpenCL), NPU (NNAPI), CPU (NEON) |
| Minimum Android | 9.0+ (API 28) |
| Architecture | arm64-v8a |
Permissions
| Permission | Why |
|---|---|
| Microphone | 录制语音用于本地转录(完全离线) |
| Internet | 仅用于下载 AI 模型,不上传任何数据 |
| Foreground Service | 在后台运行长转录任务并显示进度通知 |
| Notifications | 显示转录进度 |
Ready to experience truly offline AI?
Download LocalWhisper and let AI run on your device, not on someone else's server.