KeyUp Voice

Русская версия

⬇ Download KeyUp Voice for Windows

Latest stable installer for Windows 11 x64.

KeyUp Voice is a local push-to-talk voice typing application for Windows 11. Hold a configurable key, speak, and release it to insert recognized text into the currently focused text field.

Recognition runs locally with faster-whisper. A second hotkey can translate speech directly to English using Whisper's built-in translation mode.

Features

• Voice typing in any application that accepts pasted text.
• Separate keys, key combinations, or mouse side buttons for dictation and translation (Ctrl+Space, Alt plus the grave-accent key, Mouse 4/5, and more).
• Press Esc to cancel the current recording without recognition or insertion.
• Undo the latest insertion from the tray menu.
• Modern animated microphone indicator with seven visual styles.
• Five multilingual Whisper models selectable at first launch or later.
• View model disk usage and remove unused models or incomplete downloads.
• Automatic NVIDIA GPU detection and optional CUDA acceleration.
• CPU fallback for AMD, Intel, and systems without supported NVIDIA CUDA.
• Russian and English application and installer interfaces.
• Optional Windows startup entry.
• Resumable component downloads with size and SHA-256 verification.
• No saved voice recordings: audio is processed in memory and discarded.

Download and installation

1. Open the repository's Releases page.
2. Download KeyUp-Voice-Setup-<version>.exe.
3. Choose Russian or English in the installer.
4. Start KeyUp Voice and select a Whisper model.
5. If required, allow the application to download the model and NVIDIA CUDA runtime files.

The installer is intentionally small and does not embed Whisper or CUDA. Recognition works offline after the required components have been downloaded.

Usage

1. Place the cursor in a text field.
2. Hold the dictation hotkey and speak.
3. Release the key.
4. KeyUp Voice recognizes the speech and pastes the result.

Default hotkeys:

Action	Default
Dictation	Right Alt
Translate speech to English	Right Ctrl

Both keys can be changed in Settings from the system tray menu. Click an assignment field, then press the desired key, combination, or mouse side button. During recording, Esc cancels dictation. Undo for the latest insertion is available from the tray menu.

Whisper models

Model	Approximate download	Intended use
Tiny	75 MB	Minimum size and fastest CPU processing
Base	141 MB	Lightweight everyday use
Small	464 MB	Better accuracy with moderate resource use
Medium	1.4 GB	Recommended quality/speed balance
Large-v3	2.9 GB	Highest available accuracy

Installed models are kept separately. Switching to an installed model does not download it again. Missing models are downloaded after the selection is saved. Open Settings → Manage models to see actual disk usage and remove any model that is not active, including an incomplete download.

Model files come from the official Systran faster-whisper collection and are pinned to specific revisions.

GPU and CPU processing

• NVIDIA: KeyUp Voice can download the required CUDA 12 libraries and use float16 inference.
• AMD / Intel / no compatible GPU: processing falls back to CPU with int8 compute.

CUDA libraries are downloaded from the official NVIDIA packages on PyPI.

Privacy

• Recognition and translation run locally.
• Recorded audio is held in memory only for the current request.
• Audio is discarded after processing and is not written to disk.
• Network access is used only to download selected models and optional CUDA components.
• Recognized text is placed on the clipboard temporarily; the previous clipboard contents are restored after insertion.

See docs/PRIVACY.md for details.

Build from source

Requirements:

• Windows 11 x64
• Python 3.11 available in PATH
• PowerShell
• Inno Setup 6 to build the installer

.\setup.ps1
.\run.ps1

Build the desktop application:

.\build.ps1

Build the thin installer:

.\build-installer.ps1 -BuildApp

The result is written to installer-output. Build directories, downloaded models, CUDA files, local configuration, and installer binaries are excluded from Git.

More details are available in docs/BUILDING.md and docs/ARCHITECTURE.md.

Data locations

Data	Location
Settings and logs	%APPDATA%\KeyUp Voice
Application and downloaded components	%LOCALAPPDATA%\Programs\KeyUp Voice

Upgraded installations may retain the legacy Programs\Golos directory while displaying the current KeyUp Voice name everywhere in the user interface.

Known limitations

• A selected hotkey is reserved while KeyUp Voice is running.
• Pasting into an elevated application may fail when KeyUp Voice is running without administrator privileges.
• Whisper's built-in translation mode translates speech only to English.
• Large models require more memory and take longer to load.

Contributing and security

See CONTRIBUTING.md before opening a pull request. Please report security issues according to SECURITY.md.

License

KeyUp Voice is released under the MIT License. Third-party projects and model licenses are listed in THIRD_PARTY_NOTICES.md.

Copyright © 2026 LexGilden.