Windows · Local · GPU

Voice input for any window

Hold a global hotkey, speak Korean or English (or both), and text is typed into the focused app. Runs on-device with Whisper Large-v3 via CTranslate2 — no network round-trip.

Ctrl + Shift + Space
Default hotkey · try the demo below

Browser playground

Uses Web Speech API in-browser; the desktop app uses local Whisper instead

Ready
LARGE-V3
notes.txt
# Record or simulate — text appears here

What it does

01

Local Whisper Large-v3

faster-whisper + CTranslate2 on GPU. Korean/English code-switching without sending audio off-machine.

02

Global hotkey

Ctrl+Shift+Space works from VS Code, elevated terminals, browsers, Discord — any focused window.

03

UAC auto-elevation

Starts elevated so Unicode SendInput reaches Administrator prompts without dropping keystrokes.

04

Silence hysteresis

8s window to start speaking, then 1.8s trailing silence ends the utterance — tuned for natural pauses.

05

Non-activating HUD

PySide6 overlay with a live level meter. WS_EX_NOACTIVATE keeps focus on the target app while typing.

06

Transcribe / Translate

Tray toggle between as-spoken transcription and Korean speech → English text translation.

Quick start

PowerShell pwsh
PS ~\stt-whisper
1 Clone repository
PS > git clone https://github.com/savvy773/stt-whisper.git
PS > cd stt-whisper
2 Install dependencies
PS > uv sync
3 Launch tray app (UAC elevation)
PS > pythonw whisper-stt.pyw