Canonical has made Myna available today as a technology preview, letting Ubuntu users dictate text into the application they have open through a keyboard shortcut. The software performs speech recognition locally and can transcribe without an internet connection once the user has downloaded a model. Users can try it on Ubuntu 24.04 LTS, Ubuntu 26.04 LTS or Ubuntu 26.10.
The preview requires GNOME on Wayland and supports amd64 and amd64v3. Canonical describes Myna as an optional feature for Ubuntu 26.10, but the preview also reaches users of the two supported LTS releases. It remains under development, and Canonical cautions users to expect rough edges while testing the dictation experience.
To install it, users add the canonical-desktop-team/myna PPA with sudo add-apt-repository, then install the myna-config package through sudo apt install. They can then find Myna settings in Activities and complete the setup process before dictating. After setup, dictation starts by placing the cursor inside a text field and pressing the configured hotkey before speaking.
Canonical says Myna neither uploads nor stores the user's audio, with recognition handled entirely by the user's machine. Myna also provides visual feedback in real time while recognizing speech. Its current scope is dictation: it does not act as a voice assistant or let users operate the desktop by speaking commands.
The available recognition engines include Whisper, FunASR and Parakeet, with differences in supported languages, recognition accuracy and the hardware they require. A supported NVIDIA GPU can enable an engine with GPU acceleration. Canonical plans support for other accelerators as development continues.
Canonical is seeking tests in people's everyday applications, using their native language or another language spoken with their natural accent. It wants feedback about recognition accuracy and response speed, as well as how the software handles accents and works with applications. Testing with different microphones and computers will help it assess model behavior and default settings across Ubuntu systems.
The company also wants users to flag difficulties during installation and initial setup, along with unexpected behavior during use. It is particularly seeking accounts from people who depend on dictation, including details of capabilities they still need. Myna's development takes place openly in Canonical's myna repository on GitHub.
