16, Nov 2026 RSS feed ↗

Ubuntu Myna Can Take Dictation on the LTS Desktop You Already Use

If your daily desktop is still on Ubuntu 24.04 LTS, you can try Myna without moving it to 26.10. Canonical put local AI for accessibility on the November 12 Ubuntu Summit programme. Its separately published preview instructions list three compatible releases: 24.04 LTS, 26.04 LTS and 26.10. The detail to check before you install is how you logged in.

The desktop session is part of the requirement

Canonical’s October 12 preview announcement lists amd64 and amd64v3 and specifies GNOME on Wayland. A machine can be on the right Ubuntu release and still be running an Xorg session for an older application. Run printenv XDG_SESSION_TYPE in the session where you plan to dictate. If it reports x11, log into a Wayland session if one is available and verify it there before blaming Myna.

On a supported setup, Canonical supplies the preview through the desktop team’s Myna PPA. Its instructions add that repository and install myna-config. Adding a PPA changes the sources your package manager can use, so record it if this is a managed computer. LinuxHospital’s guide to checking an installed APT package revision is a useful reminder to distinguish a repository’s announcement from the packages actually installed on your system.

Try the applications that will receive the words

Open Myna settings from Activities and follow setup to download a recognition model. Then put the cursor in a text field, invoke the configured shortcut and dictate several sentences. This is a practical compatibility check: getting a transcription in one text editor tells you that editor accepts the result, while your browser or work application may still be worth checking separately. Include punctuation, names and a few terms you actually use.

Canonical says transcription runs locally after the model download and audio is neither uploaded nor stored. That gives you a straightforward second check on a spare machine: once setup is finished, disconnect it from the network and dictate into an offline editor. If setup itself is still waiting for a model, an offline attempt at that point tests a different thing.

Myna offers Parakeet, Whisper and FunASR. Canonical flags trade-offs in language coverage, accuracy and hardware. If the first model struggles with your language or microphone, keep the same short passage when trying another engine. The preview can offer GPU acceleration on supported NVIDIA hardware, but the presence of an NVIDIA card alone does not tell you whether that option will appear.

A small pilot can answer two separate questions. First, does the shortcut put words into the application you need? Second, does it keep doing so when that computer is offline? Keep the same model and passage for the second check so the network is the only thing you change. Then try the language and microphone you would use every day.

C

Chris Page

Chris Page covers software, connected devices and the businesses behind digital services, with a reporting style built around the details that change a reader's decision. His journalism studies at the University of Missouri provide the background to a methodical interest in documents, release histories and the difference between a confirmed result and an expectation. Chris contributes to the group's technology, gaming and music titles, moving between a Linux upgrade question, an Apple product development and the economics of an online entertainment platform. He tends to organise a story around what changed, who is affected and what remains unresolved. Clear comparisons and specific limitations are recurring features of his work, especially when an announcement sounds simpler than the underlying terms or technical requirements.

More from this author →