VoicePi is a macOS 14+ menu-bar voice input app built with Swift Package Manager.
It lets you press a keyboard trigger to start recording speech, transcribe audio with either Apple Speech or a remote ASR provider (OpenAI-compatible, Aliyun ASR, or Volcengine ASR), optionally refine the transcript with an OpenAI-compatible LLM, and then paste the final text into the currently focused input field when you press the trigger again.
Features
- Menu-bar only app (
LSUIElement) — no Dock icon - Apple Speech local transcription support
- Remote large-model ASR support (OpenAI-compatible, Aliyun ASR, and Volcengine ASR)
- Real-time floating capsule overlay with live waveform
- Clipboard-based paste injection with automatic clipboard restore
- Optional LLM refinement for conservative ASR correction
- Support for multiple recognition languages (Chinese, English, Japanese, Korean)
- Configurable shortcuts and triggers
Installation
VoicePi can be installed through Homebrew:
brew tap pi-dal/voicepi https://github.com/pi-dal/VoicePi
brew install --cask pi-dal/voicepi/voicepi
Or download directly from GitHub Releases.
Usage
- Launch VoicePi — a microphone icon appears in the menu bar
- Press the configured hotkey trigger to start recording
- Speak
- Press the trigger again to stop recording and paste the transcribed text
While recording, a floating capsule appears with live waveform and streaming transcript.
Open Source
VoicePi is released under the MIT License.
