System-wide, privacy-first dictation

Dictate anywhere. Text lands at your cursor.

Keep Viska in the background while you write in email, chat, documents or code. Press a global hotkey, speak naturally, and Viska transcribes locally and inserts the finished text at your cursor.

Viska 0.5.60 delivers background dictation only to its captured external target and adds a session-only list of recent transcription segments.

Viska desktop app screenshot showing the clean Qt interface with record button, transcription editor and example text demonstrating local model transcription for privacy
Runs in the background Global hotkeys Automatic insertion Local Whisper

Viska app download

Linux first. Desktop-ready beyond Linux.

Viska is designed and tested first for Kubuntu/KDE Plasma. The same local-first app is now also available as a macOS Apple Silicon preview. A Windows preview remains intentionally unpublished until hands-on testing is complete.

Linux · Primary

Debian package

For Kubuntu, Ubuntu and Debian-family systems. Includes verified KDE Plasma clipboard synchronization, model preloading and an optional guided NVIDIA accelerator with larger CT2 models—without changing your GPU driver.

Download .deb Permanent link: viska-latest.deb
Version
0.5.60
Size
347 KiB
SHA256
e8e5097982879f4b98ef6de9e1b635c2d883c68073e0f39a426a080f88c3d964

macOS · Apple Silicon preview

macOS app bundle

For Apple Silicon Macs. Double-press left Command to dictate from any app, then let Viska insert the finished text at the cursor. Input Monitoring enables the hotkey and Accessibility enables automatic insertion. This unsigned preview may require Control-click → Open on first launch.

Install FFmpeg with brew install ffmpeg to import audio files and export OGG.

Download for macOS Permanent link: viska-latest-macos-arm64.zip
Version
0.5.60
Size
96 MiB
SHA256
7ece46d95a10035fb337578fc1a04ae6bb74b96ccc7d7715963cba72505aaf7c

Windows · In development

Windows package

A Windows installer is not published yet. It will be added here after a native build and verification, not as an untested placeholder.

Viska family

Local-first tools for writing and speaking on your desktop.

Viska

Viska

Dictate with local Whisper models, hotkeys and optional cloud engines. Linux/KDE Plasma stays the primary experience.

Open downloads
Viska Brev

Viska Brev

Proofread Thunderbird drafts before you send them, locally through Ollama or through OpenRouter with your own key.

Open Viska Brev page

How it works

Stay in your app. Viska comes to you.

1

Place the cursor

Stay in the email, chat, document or prompt where you want the words to appear.

2

Press the global hotkey

Start and stop dictation without opening or switching to the Viska window.

3

Speak and continue

Viska transcribes, optionally proofreads, and inserts the finished text at your cursor. If desktop integration is unavailable, the text is safely copied to the clipboard instead.

Viska Brev

Your email, proofread before you press send.

Viska Brev is a Thunderbird 128+ extension for Swedish and English email drafts. It keeps the same local-first philosophy as Viska: use Ollama on your computer, or connect OpenRouter only when you choose cloud proofreading.

Spelling Grammar Expand Diff preview Quotes and signatures stay untouched Ctrl+Alt+P / 1 / 2 / 3
Viska Brev in action inside Thunderbird: the proofreading panel shows a real diff with red/green highlights on text, '13 ändringar', and buttons Ersätt / Förkasta.

Thunderbird extension

Install Viska Brev from an XPI file.

Thunderbird allows installing unsigned extensions from a local file. Download the XPI, verify it against SHA256SUMS if needed, and install it from Thunderbird's Add-ons and Themes screen.

Download Viska Brev .xpi viska-brev-0.3.1.xpi
Version
0.3.1
Size
49 KB
Requires
Thunderbird 128+
SHA256
465bb31ca3fe758d728dabf757d788d8b53e9aad888e0b2a7127d46a9f92e310

Viska Brev setup

Add local or cloud proofreading to Thunderbird.

Install the XPI from Thunderbird, then choose whether proofreading runs locally through Ollama or through OpenRouter with your own API key.

1. Download viska-brev-latest.xpi
2. Thunderbird -> Tools -> Add-ons and Themes
3. Gear menu -> Install Add-on From File
4. Local AI: ollama pull gemma3
5. Cloud AI: add your OpenRouter API key

Install

Download, verify and install.

Use the checksum before installing. Package installation can pull Python dependencies from PyPI during setup.

wget https://getviska.app/downloads/viska_0.5.60_all.deb
sha256sum viska_0.5.60_all.deb
sudo apt install ./viska_0.5.60_all.deb

What it does

A focused dictation tool for your desktop.

Works where you work

Leave Viska in the background and dictate directly into other applications with global hotkeys and automatic text insertion.

Local by default

Run speech-to-text on your machine with curated Whisper models, including Swedish-friendly options, and save the latest recording as a WAV file when you need the audio.

KDE and Wayland aware

Built around portal hotkeys, Linux text-insertion backends and a visible clipboard fallback, so a missing desktop capability never silently loses your words.

Proofreading

Choose Off, private local proofreading through Ollama, or an explicitly selected cloud provider before Viska inserts the finished text.

Optional NVIDIA acceleration

On Debian-family Linux, Viska verifies CTranslate2/CUDA and guides you through a private faster-whisper runtime. GPU-ready choices now include Swedish KB-Whisper Medium and Large plus multilingual Whisper Large V3 and the faster Large V3 Turbo. Green GPU · CT2 badges identify the accelerated format; your distribution remains responsible for the NVIDIA driver.

Cloud models, desktop flow

Use providers like Deepgram, OpenAI, OpenRouter, xAI, Gemini and Mistral while keeping Viska's desktop hotkeys and UI.

Privacy-first

Audio stays local unless you explicitly choose a cloud provider. No telemetry is collected.

Complementary tools

Viska transcribes. Viska Brev proofreads.

Viska turns your speech into text. Viska Brev proofreads and polishes your writing. Use the two applications together — in parallel — for a complete local-first workflow on Linux. A future bridge will connect them even tighter via Viska’s D-Bus API.

Compatibility

Built for Kubuntu first.

Changelog

Recent development.

  1. 0.5.60

    Routes each background dictation exactly once to its captured external target, leaves the Viska editor unchanged and adds a session-only list of recent transcription segments.

  2. 0.5.49

    Makes double-pressing the left Command/Meta key start or stop dictation on macOS, with Command-Shift-D retained as a fallback. Completed dictation continues to copy to the clipboard by default.

  3. 0.5.48

    Publishes a smaller Apple Silicon macOS preview bundle with built-in silence-detection fallback.

  4. 0.5.47

    Introduces the first macOS preview, a native global-hotkey backend and a verified Flatpak packaging route.

  5. 0.5.35

    Improves long local Whisper transcription with bounded overlapping chunks and useful performance logging.

  6. 0.5.34

    Fixes audio-file import by opening KDE's native kdialog picker first (with fallbacks).

  7. 0.5.33

    Uses the Linux desktop FileChooser portal so KDE Plasma owns the audio import picker.

  8. 0.5.32

    Moves audio-file import to native chooser while keeping themed picker for other tasks.

  9. 0.5.31

    Improves model download progress with measured MB and stable scrolling.

Full changelog (all versions)

FAQ

Before you install.

Is this stable?

No. Viska is still a fast-moving test build. Expect field-test fixes and frequent releases.

Does press-to-talk work on Wayland?

On modern KDE Plasma it should work through the XDG GlobalShortcuts portal. Other desktops may need fallbacks.

Why host a .deb directly?

It is the fastest path for early testing. A signed APT repository is the better long-term distribution path.

Why host the Thunderbird extension directly?

Viska Brev is not on addons.thunderbird.net yet. Thunderbird can install the self-hosted XPI from a local file.