Ubuntu 26.10 finally has built-in speech-to-text, running completely locally. macOS has Voice Control, Windows has Voice Access. Linux desktop has always been a vacuum in this area.

I'm not saying there are no speech-to-text tools on Linux—Speech Note, running Whisper yourself, they all work.

💡 What You Will Learn

Ubuntu 26.10 finally has built-in speech-to-text, running completely locally. macOS has Voice Control, Windows has Voice Access. Linux desktop has always been a vacuum in this area. I'm not saying

📜 Table of Contents

• 01. What is Myna?
• 02. Privacy First, Not a PPT Slide
• 03. Architecture: Modular, GNOME First
• 04. Don't Overthink It—What Myna Is Not
• 05. What's the Community Saying?
• 06. What Does This Mean for Regular Users?
• 07. Personal Take: This Matters More Than It Seems
• 08. What Do You Think?

Ubuntu 26.10 Finally Gets Built-in Speech-to-Text, Fully Local

macOS has Voice Control, Windows has Voice Access. Linux desktop has been a vacuum in this area.

I'm not saying there are no speech-to-text tools on Linux—Speech Note, running Whisper yourself, they all work. But a system-level, native experience where you press a hotkey and it just works? That hasn't existed.

Now Canonical is stepping up.

01. What is Myna?

The project is codenamed Myna, after the myna bird—known for mimicking human speech.

Core functionality in one sentence: Press a hotkey, speak, and text appears in whatever app you're using.

Ubuntu 26.10 "Stonking Stingray" (expected October 2026) will ship it for the first time.

It's not some flashy AI voice assistant. It's straightforward dictation—speech-to-text, fully local.

Project lead jibel said on Ubuntu Discourse:

"For Ubuntu 26.10, we're deliberately focusing on the basics: a reliable desktop dictation."

Get dictation right first, then talk about everything else.

02. Privacy First, Not a PPT Slide

What reassures me most about this: Privacy is a design principle, not a patch added later.

Here's what they're doing:

• Speech recognition model runs locally, no internet required
• Microphone only activates when you manually start dictation
• Audio is processed in memory and discarded immediately, no recordings uploaded
• No network needed—once the model is installed, it works offline

Same logic as macOS's local dictation and Windows' Voice Access. Your voice never leaves your machine.

In today's "AI must go to the cloud" era, this is a breath of fresh air.

03. Architecture: Modular, GNOME First

Myna's architecture has three layers: speech recognition engine, dictation manager, and text injection layer.

The benefit of this design: Upgrading one component in the future won't require touching the whole system.

Initially, it only supports Wayland + GNOME. The logic is simple—Ubuntu's default is GNOME, so get that stable first, then support KDE, Budgie, and other desktop environments.

From an engineering standpoint, this is the right call. Don't try to cover everyone from the start; let the default users get a working experience first.

04. Don't Overthink It—What Myna Is Not

The official team explicitly listed what this version won't do:

• ❌ Voice assistant (no "Hey Myna" stuff)
• ❌ Voice control of the desktop (saying "open browser" won't work)
• ❌ Translation or automatic language detection
• ❌ Complex scenarios

The first version does one thing: Press hotkey → Speak → Text appears.

This restraint—honestly—I really appreciate it. Too many projects in the Linux ecosystem try to "build a big platform" and never finish. Myna locking down its scope first makes it feel like something that can actually ship.

05. What's the Community Saying?

Reactions on Reddit have been fairly positive.

On the Ubuntu subreddit, user CobaltOne said:

"I've been looking for months. A system-level built-in feature—this is awesome."

Of course, there are voices wondering if "Canonical's previous AI plans are going to fizzle out again"—after all, Canonical made a big deal about AI at the Ubuntu Summit, and everyone is waiting to see actual delivery.

The GitHub repo canonical/myna currently has 33 Stars, 1 Fork, under GPL-3.0. Code hasn't been released much yet—only architecture docs and planning notes. But the repo was created on June 5th, just two weeks ago, so early stage is expected.

06. What Does This Mean for Regular Users?

If you're a Linux desktop user, here's the direct value:

• Taking notes, writing articles—dictation is 2-3x faster than typing
• For people with disabilities or hand injuries, system-level dictation is a necessity
• No need to install third-party software or configure API keys for voice input
• Everything offline, no subscription fees, no privacy concerns

Don't underestimate that last point. Speech-to-text SaaS products (Otter.ai, etc.) cost at least a dozen dollars a month. Myna is open-source and free.

07. Personal Take: This Matters More Than It Seems

Ubuntu's speech-to-text isn't technically groundbreaking—Whisper and Speech Note have been running in the community for a while. But from an ecosystem perspective, it's a signal.

First, Linux desktop is filling in the "exists but isn't polished" gaps. HiDPI scaling took a decade to stabilize, HDR only started rolling out last year. Basic capabilities like voice input—Windows and macOS users take for granted—Linux users had to DIY. Now Canonical is investing officially, which shows they're taking desktop experience seriously.

Second, the local-first direction. The hottest buzzwords in AI right now are "Agent," "multimodal," "cloud inference." Canonical chose the most unassuming thing—speech-to-text—and emphasized local, no-data-upload. In the middle of the LLM hype wave, this feels old-school. But for users, a dictation tool that doesn't need internet, doesn't burn GPU, and just works reliably is more practical than 100 cloud AI agents.

Third, don't get too excited too soon. Honestly, Canonical has drawn up plans for desktop features before—Ubuntu Phone, Mir display server, Unity 8—none of them succeeded. Whether Myna ships on time with Ubuntu 26.10 and delivers a stable experience—I hope it does, but I'll wait until the first beta is running before I get my hopes up.

08. What Do You Think?

Native speech input on Linux desktop—finally here after all these years. Do you think Canonical can pull it off this time? Or is it another "shipped but nobody uses" feature?

How important do you think speech-to-text is for Linux desktop adoption? Drop your thoughts in the comments.