Press a key and speak to insert text locally
When typing on a keyboard is cumbersome, hold a specific key and speak; the content is immediately converted to text and pasted into the current application. Hex processes this entire workflow on your local computer, so you do not need to worry about internet connectivity or sending data to external servers. The voice model is downloaded only on the first run, and after that, data does not leave your device.
Input flow starting with a shortcut
The usage is very simple. Hold the designated shortcut key while speaking, then release it. Hex uses its built-in speech recognition model to convert sound to text and automatically pastes the result into the application that currently has focus. Experimental features can be found in the 'Labs' section at the bottom of the sidebar.
Installation paths by platform
It is available on macOS, Windows, and Linux, with installation files provided for each platform. macOS offers a DMG file for Apple Silicon, Windows provides installers for x64 and ARM64, and Linux supports x86_64 and ARM64 architectures. Linux users can choose between .deb, .rpm, or executable files depending on their distribution. Installation via package managers is also possible, utilizing Homebrew, AUR, Nix, or the official installation script.
Version selection and contribution methods
The Stable version refers to the current release, while the Dev version refers to the latest build from the dev channel, which updates automatically from that channel. Choose Stable if you prioritize stability, or the Dev channel if you want the latest features. This project is released under the MIT license but does not accept pull requests. If you have a bug or a new idea, leave a description in an issue, and the team's agents will handle it quickly.
Pre-installation checkpoints
Because Hex performs speech recognition locally, a model download is required on the initial run. Once this process is complete, it works without an internet connection, though disk space may be consumed depending on the model size. For certain Linux distributions, the installation method may vary based on package manager support, so it is best to check the file format suitable for your environment.
