Speak. Transcribe. Paste.
English | 简体中文
Whisp is a desktop speech-to-text app focused on one thing: instantly turning your voice into text and pasting it where your cursor is.
Powered by OpenAI-compatible transcription APIs (OpenAI by default), with preset model suggestions and custom model support. Built with Tauri v2.
- Check for Updates — Built-in update checker that queries GitHub Releases for new versions, shows release notes, and provides direct download links.
- Improved Toggle Descriptions — Settings toggle rows now display descriptive text below each label for better clarity.
- API Key Persistence Fix — Fixed a critical bug where API key was not saved to disk, causing "not configured" prompt on every restart.
- UTF-8 Safety — Fixed a potential crash when truncating multi-byte (CJK) error messages.
- Performance — Reduced memory allocations in the transcription retry loop.
- Silence auto-stop — Configurable silence detection automatically stops recording when you stop speaking, no need to press the hotkey again.
- Whisper prompt support — Add context hints to improve transcription accuracy for domain-specific vocabulary.
- Silence threshold tuning — Adjust the silence detection sensitivity to match your environment.
- Japanese UI — Full Japanese language support alongside Chinese and English.
- Mic selection — Choose your preferred microphone device from settings.
- Overlay position memory — The waveform overlay remembers its last position across sessions.
- Launch at startup — Option to automatically start Whisp when you log in.
- History pagination & batch operations — Paginated history loading with batch delete and clear all.
- Multi-provider auto-detection — Automatically identifies OpenAI, Groq, Fireworks, Deepgram, Google Cloud, or Custom providers.
- Tray icon support — System tray with custom template icons for macOS dark/light mode.
- API keys are now stored in the system keychain instead of plaintext settings.
- Transcription requests now support configurable timeout and automatic retries.
- Silence trimming reduces upload size and improves responsiveness.
- History now tracks success / failure state, provider, language, and retry-ready audio availability.
- Settings now include auto-paste, audio retention, paste delay, and reliability controls.
Whisp is made for people who type a lot but think faster than they can type:
- Press once, speak naturally.
- Press again, get accurate text.
- Text is auto-pasted to the current app.
No complex workflows, no heavy setup, no context switching.
- Tap
Right ⌘on macOS /Right Ctrlon Windows (customizable) - Speak
- Tap again to stop (or let silence auto-stop kick in) — text is transcribed and pasted instantly
- One Shortcut — Global hotkey to start/stop recording. No UI to navigate.
- Auto-Paste — Transcribed text goes straight to your cursor. No copy needed.
- Silence Auto-Stop — Automatically stops recording after configurable silence duration.
- Model Presets + Custom Models — Built-in popular model names plus free-form custom model input.
- Model Guide — In-app model guide button with model descriptions and selection hints.
- Waveform Overlay — Minimal always-on-top visualizer while recording, position persists across sessions.
- History — Rich local history with success/failure status, search, pagination, batch delete, and retry support.
- Multi-Language UI — Simplified Chinese, English, and Japanese interface.
- System Tray — Runs quietly in the background with launch-at-startup option.
- Check for Updates — Built-in update checker with release notes and direct download links.
Download the latest DMG for macOS from Releases:
| File | Architecture |
|---|---|
Whisp_2.5.1_aarch64.dmg |
Apple Silicon (M1/M2/M3/M4) |
Whisp_2.5.1_x64.dmg |
Intel Mac |
真的会爱上这种"张嘴就能写字"的效率感 ✨
Whisp 就是那种你用了就回不去的办公小工具：
- 开会复盘：边听边说重点，秒变文字。
- 写作卡壳：先说出来，再慢慢润色。
- 日常回复：不用来回切输入法，想到就说。
一句话总结：把"打字焦虑"换成"说话自由"。
Prerequisites: Node.js and Rust.
git clone https://github.com/sexyfeifan/Whisp.git
cd Whisp
npm install
npm run tauri dev
Speak. Transcribe. Paste.
© 2026 sexyfeifan