Local Windows voice tools

Admix SpeakDeck

Read selected text. Hold to dictate.

Admix SpeakDeck gives one Stream Deck action two Windows workflows: tap to read the current selection aloud, or hold to dictate and type into the application that originally had focus.

Double-tap stops active speech or dictation. The English recognition model and required runtimes are bundled, and no Admix account is required.

  • Local speech recognition
  • One key, three gestures
  • No Admix account
Admix SpeakDeck key icon in its idle state.
The idle state on a Stream Deck key.
Product maturity
Complete and working
The plugin is complete and working on its documented Windows platform.
Public availability
Ready for release
A public Marketplace or package release is not available yet.

What it does

Built around the real workflow.

Tap to read

  • Tap to read the current selected text through an installed Windows SAPI voice.
  • A new request replaces active speech instead of overlapping it.
  • Double-tap to stop active speech or dictation.

Hold to dictate

  • Hold through the ready state, speak, then release to transcribe locally.
  • Text returns to the application that originally had focus.
  • If focus changes during processing, SpeakDeck refuses to type into the new window.

Local and self-contained

  • The bundled English Whisper model runs locally.
  • Normal use requires no separate Python, Whisper, CUDA, Vulkan SDK, or Visual C++ runtime installation.
  • The plugin itself makes no network requests during normal use. A user-selected third-party Windows voice may use its own network service.

Actual product assets

Requirements and boundaries

What it needs.

  • Windows 11 x64
  • Elgato Stream Deck 7.1 or later
  • An enabled Windows SAPI voice
  • Desktop microphone permission for dictation
  • A Vulkan-capable GPU is recommended; bundled CPU fallbacks are available
  • The target application must not be elevated above Stream Deck if typed input is expected

Why it exists

The problem behind the product.

SpeakDeck started with a simple problem: reading selected text and dictating a reply should not require a pile of disconnected tools. The result is one Stream Deck action with gestures that stay out of the way once they become muscle memory.

A focused product

Not publicly available yet.

Public install, download, or application links will appear here when they are available.