What is Hex?
Hex is a local voice dictation app for macOS, Windows and Linux, published by Anomaly under the anomalyco GitHub organisation. You hold a keyboard shortcut, speak and let go; Hex transcribes on your device and pastes the text into whatever app you are using. The first run downloads a speech model, and after that nothing leaves your machine.
Hex recommends a speech model for the system's preferred language, such as Parakeet Unified English for English, and you can add your own GGUF model file or reuse models another app already downloaded. A vocabulary list biases transcription toward names you supply. macOS, Windows and Linux share one window with Settings, Modes and History, Meetings are macOS-only, and unfinished features sit behind their own switches in Labs. Agents can read retained transcripts and control the running app through the release CLI, and a public TypeScript transcription SDK is available.
Hex is free and MIT licensed. On macOS it requires Apple silicon and macOS 15 or later and installs from a DMG or a Homebrew cask. Windows x64 and ARM64 and Linux x86_64 and ARM64 builds are in beta, with Windows installers, .deb and .rpm packages, a Linux install script, AUR and Nix. Stable and Dev channels are offered, and the project takes bug reports and ideas through GitHub issues rather than pull requests.
Key capabilities
- Hold-to-talk dictation Hold the shortcut, speak and let go, and Hex pastes the transcript into the app you are in.
- On-device transcription After the first speech model download, speech is transcribed locally and nothing leaves the machine.
- Model choice Hex recommends a model for your system language, and you can add your own GGUF file or reuse downloaded models.
- Custom vocabulary A vocabulary list biases transcription toward the names you list.
- History and Modes One window holds Settings, Modes and History on macOS, Windows and Linux.
- Annotate Hex records the screen while you draw and talk, then sends the recording to an agent.
- CLI control The release CLI lets agents read retained transcripts and control the running app.
Where Hex fits
- macOS app A signed build for Apple silicon on macOS 15 or later, installed by DMG or Homebrew cask, with Meetings available only on macOS.
- Windows beta Installers for x64 and ARM64, plus a PowerShell install command on the website.
- Linux beta X86_64 and ARM64 builds for X11 and wlroots-based Wayland compositors such as Hyprland or Sway, using Vulkan with a CPU fallback.
- Single user process On Linux, Hex runs as one process per user rather than a system-wide or root daemon.
- Release channels Stable is the current release, and Dev is the newest build that updates itself from the dev channel.
- Labs Every unfinished feature gets its own switch, off by default.
- Developer SDK A public TypeScript transcription SDK works with an explicitly supplied native helper.
Install
brew install --cask anomalyco/tap/hexReviews of Hex
Be the first to share your experience with Hex.
Write a review
Loading review form…
Alternatives to Hex
AmicalMIT-licensed desktop dictation app that pastes speech into any focused app, using local Whisper models offline or the hosted Amical Cloud model, with optional AI formatting, vocabulary, snippets and transcription history.
AnarlogAnarlog is an MIT-licensed AI meeting notepad for macOS, Windows and Linux. It captures meetings without a bot, transcribes on-device or with your chosen provider, and keeps notes in local SQLite with optional encrypted sync.
DograhDograh is a BSD-2-Clause voice agent platform that you can run with Docker. It combines a visual workflow builder, bring-your-own LLM, speech-to-text and text-to-speech providers, telephony integrations and an MCP server.