bopuppy101/dbdude-v2t

Local, privacy-first voice-to-text for the desktop: hold a hotkey, speak, and the text appears at the cursor.

Python

0

68 commits

updated Oct 10, 2026

See the code

README

dbdude-v2t

Website: https://dbdude-v2t.dbdude.net/

Local, privacy-first voice-to-text for your desktop. Hold a hotkey to record, release to transcribe, and text appears at your cursor - in any application.

Features

  • System-wide input - types into any focused window (browser, editor, terminal, etc.)
  • Fully local - runs Whisper models on your machine, no cloud, no API keys, your audio never leaves your computer
  • Multiple models - choose from tiny to large-v3 depending on your hardware
  • GPU accelerated - CUDA detection on Windows, X11 Ubuntu, and Omarchy; Apple Silicon (MLX) on macOS. The Ubuntu 26.04 launcher currently uses CPU
  • Custom mappings - correct misheard words, expand abbreviations, add programming symbols
  • Mapping packs - bundled Punctuation and Programmer packs for common replacements
  • System tray app - runs in the background with status icons (idle/recording/transcribing)
  • Configurator GUI - settings, model selection, audio device selection
  • Continuous mode - toggle on for long-form dictation

Dictation punctuation

All platform versions omit the automatic final period for a single sentence, regardless of its length. Multiple sentences keep punctuation and have their sentence starts capitalized. Sentence detection uses the recognizer's punctuation and accounts for common abbreviations; ambiguous abbreviations may still need correction. Question marks and exclamation marks are preserved.

Enable the Punctuation mapping pack, or add individual custom mappings, to use question mark → ?, exclamation point / bang → !, and dot → .. Spoken dot preserves an explicitly requested period even for a single sentence. Punctuation attached by the recognizer to a spoken command is replaced along with the command, so bang! produces !; bang bang still produces !!. Windows and macOS custom formatting rules continue to run after sentence formatting.

Restart the app after updating the Python code. Mapping-only edits can be loaded with Reload Mapping Files.

Platforms

PlatformHotkeyText OutputAI Engine
WindowsAlt+Shift (hold)AutoHotkeyfaster-whisper (CPU/CUDA)
macOSFn (hold)Quartz eventsmlx-whisper (Apple Silicon)
Ubuntu (X11)Alt+Shift (hold)xdotoolfaster-whisper (CPU/CUDA)
Ubuntu 26.04 (Wayland)Alt+Shift (hold)ydotoolfaster-whisper (launcher uses CPU)
Omarchy (Arch + Hyprland)Alt+Shift (hold)ydotoolfaster-whisper (CPU/CUDA)

Keyboard Shortcuts

ShortcutAction
Alt+Shift (hold/release)Record / Stop & transcribe
Ctrl+Shift+SpaceToggle continuous mode
Ctrl+Shift+QExit

macOS uses Fn instead of Alt+Shift.

Installation

Choose your platform and preferred level of detail. Each guide is self-contained.

macOS (Apple Silicon)

Windows

Ubuntu with X11/Xorg

Ubuntu 26.04 with Wayland

Omarchy (Arch Linux + Hyprland)

License

GPL-3.0-or-later - Free and open source. You may use, modify, and redistribute it, including commercially; distributed versions must also be released under the GPL.

Copyright (c) 2025-2026 Michael Foster / DBDude Inc.

bopuppy101/dbdude-v2t

Local, privacy-first voice-to-text for the desktop: hold a hotkey, speak, and the text appears at the cursor.

Python

0

68 commits

updated Oct 10, 2026

See the code

README

dbdude-v2t

Website: https://dbdude-v2t.dbdude.net/

Local, privacy-first voice-to-text for your desktop. Hold a hotkey to record, release to transcribe, and text appears at your cursor - in any application.

Features

  • System-wide input - types into any focused window (browser, editor, terminal, etc.)
  • Fully local - runs Whisper models on your machine, no cloud, no API keys, your audio never leaves your computer
  • Multiple models - choose from tiny to large-v3 depending on your hardware
  • GPU accelerated - CUDA detection on Windows, X11 Ubuntu, and Omarchy; Apple Silicon (MLX) on macOS. The Ubuntu 26.04 launcher currently uses CPU
  • Custom mappings - correct misheard words, expand abbreviations, add programming symbols
  • Mapping packs - bundled Punctuation and Programmer packs for common replacements
  • System tray app - runs in the background with status icons (idle/recording/transcribing)
  • Configurator GUI - settings, model selection, audio device selection
  • Continuous mode - toggle on for long-form dictation

Dictation punctuation

All platform versions omit the automatic final period for a single sentence, regardless of its length. Multiple sentences keep punctuation and have their sentence starts capitalized. Sentence detection uses the recognizer's punctuation and accounts for common abbreviations; ambiguous abbreviations may still need correction. Question marks and exclamation marks are preserved.

Enable the Punctuation mapping pack, or add individual custom mappings, to use question mark → ?, exclamation point / bang → !, and dot → .. Spoken dot preserves an explicitly requested period even for a single sentence. Punctuation attached by the recognizer to a spoken command is replaced along with the command, so bang! produces !; bang bang still produces !!. Windows and macOS custom formatting rules continue to run after sentence formatting.

Restart the app after updating the Python code. Mapping-only edits can be loaded with Reload Mapping Files.

Platforms

PlatformHotkeyText OutputAI Engine
WindowsAlt+Shift (hold)AutoHotkeyfaster-whisper (CPU/CUDA)
macOSFn (hold)Quartz eventsmlx-whisper (Apple Silicon)
Ubuntu (X11)Alt+Shift (hold)xdotoolfaster-whisper (CPU/CUDA)
Ubuntu 26.04 (Wayland)Alt+Shift (hold)ydotoolfaster-whisper (launcher uses CPU)
Omarchy (Arch + Hyprland)Alt+Shift (hold)ydotoolfaster-whisper (CPU/CUDA)

Keyboard Shortcuts

ShortcutAction
Alt+Shift (hold/release)Record / Stop & transcribe
Ctrl+Shift+SpaceToggle continuous mode
Ctrl+Shift+QExit

macOS uses Fn instead of Alt+Shift.

Installation

Choose your platform and preferred level of detail. Each guide is self-contained.

macOS (Apple Silicon)

Windows

Ubuntu with X11/Xorg

Ubuntu 26.04 with Wayland

Omarchy (Arch Linux + Hyprland)

License

GPL-3.0-or-later - Free and open source. You may use, modify, and redistribute it, including commercially; distributed versions must also be released under the GPL.

Copyright (c) 2025-2026 Michael Foster / DBDude Inc.