voicepaste

Global push-to-talk voice transcription tool. Press a hotkey to start recording, press again to stop. Audio is transcribed and pasted into the currently focused element.

How it works

1. Ctrl+Shift+Space → start recording from microphone
2. Ctrl+Shift+Space again → stop recording
3. Audio is saved to a temp WAV file and sent for transcription
4. Transcribed text is pasted into whatever is focused (clipboard + Ctrl+V)

Prerequisites

Windows

• PortAudio: download from http://www.portaudio.com or install via vcpkg/msys2
• No additional tools needed (uses Win32 APIs for hotkey and paste)

Linux (Ubuntu/Debian)

sudo apt install libportaudio2 portaudio19-dev libx11-dev xclip xdotool

Building (both platforms need Go 1.22+)

# Native build (auto-detects platform)
make build

# Or explicitly:
make build-linux
make build-windows  # needs mingw: sudo apt install gcc-mingw-w64-x86-64

Usage

./voicepaste

Then press Ctrl+Shift+Space to toggle recording. Logs go to stderr.

WSL2

Run the Windows binary. It captures the global hotkey, records via Windows audio, and pastes into whatever window is focused — including your WSL2 terminal.

Architecture

cmd/voicepaste/main.go       — entry point, wires everything together
internal/
  hotkey/
    hotkey.go                 — shared Handle type
    hotkey_windows.go         — Win32 RegisterHotKey + GetMessage loop
    hotkey_linux.go           — X11 XGrabKey + XNextEvent loop
  audio/
    audio.go                  — PortAudio recording + WAV encoding
  paste/
    paste_windows.go          — Win32 clipboard + SendInput(Ctrl+V)
    paste_linux.go            — xclip + xdotool
  transcribe/
    transcribe.go             — stub (returns "ABCXYZ"), replace with API call

TODO

• Replace transcribe stub with OpenAI Whisper API call
• Config file for hotkey, API key, etc.
• Audio level indicator / tray icon
• Wayland support (wl-copy + wtype/ydotool)
• Optional: hold-to-talk mode