版本说明

🎙️ Dictation without an account, on your own machine

Vito can now install and run a speech engine itself. Pick Vito local (Parakeet) in Settings → Speech recognition, press install once, and that is the whole setup: no provider account, no API key, no server of your own, and no audio leaving your computer.

Behind it is parakeet.cpp with NVIDIA's Parakeet TDT 0.6B v3 — 25 European languages, detected automatically. Vito downloads a pinned build and the model (≈ 900 MB, once), checks both against hashes baked into the release, and runs the engine as a background process only while you have that card selected.

Measured on a 12-core desktop:

It needs about 1 GB of RAM while it runs and 0.9 GB of disk. There is no live transcript — the text arrives when you stop speaking. Keyterms are not used by this engine; your corrections still are.

🔌 Or point Vito at any OpenAI-compatible endpoint

The other new provider takes a base URL and talks to whatever is behind it: whisper-server, Speaches, LocalAI on your own machine, or hosted Whisper at Groq and OpenAI.

Whisper always encodes 30 seconds of audio, so a short dictation costs the same as a long one. Where whisper.cpp may be listening, Vito now sizes the encoder window to the recording — that took a 13-second clip from 9.0 s to 4.1 s on CPU. It also finds whisper-server on its own /inference path, and joins whisper.cpp's segments back into one line instead of typing its newlines into your document.

✍️ Cleanup rules you can write yourself

The cleanup prompt is the biggest lever on what comes out, so it is yours now. Five built-in sets — Vito default, Verbatim, Business writing, Notes, Messages — are always in the list and read-only; start your own from any of them.

Only the *rules* are yours: Vito appends a fixed output contract to whichever set is active, so a rule set can change what cleanup does and can never make it paste commentary at your cursor.

New Thinking control alongside it, for endpoints that support reasoning_effort. Turning it off is the fix when a model spends thousands of tokens deliberating and returns nothing — raising the token budget never was.

🧾 Why a dictation came out unpolished

When the AI pass fails, the raw transcript is pasted and a toast disappears seconds later. The reason is now kept on the history entry, so you can still find out hours afterwards. The notification points you there instead of truncating a provider error into uselessness.

The out-of-credit card can be dismissed, and stays gone until that account actually runs dry again.

🎛️ The speech settings, rearranged

The page was one long list of every field of every provider. Each card now carries its own fold-out — open for the provider you use, closed for the rest — with the provider's logo and one line saying what it asks of you: an account somewhere, a server of your own, or nothing but a download.

🐛 Fixes

• A dropped streaming connection fell back to AssemblyAI's async API whatever provider you had chosen. A Soniox user without an AssemblyAI key simply lost the dictation.
• The silence watchdog and auto-stop read speech activity off the live transcript, which a batch provider never sends. They now fall back to the microphone level, so both work with the new providers.

🌍 Translations

All of the above is translated in all 60 languages.

Install

• Windows — Vito-Setup-2026.9.exe (per-user, no admin).
• macOS — Vito-2026.9.dmg — drag Vito to Applications. First launch needs right-click → Open, then two permissions (Microphone, Accessibility).
• Linux — Vito-2026.9-x86_64.AppImage — chmod +x and run — or the Flatpak from vinceecniv.github.io/Vito.

Checksums (SHA-256)

b9ed8de7e58fb36c18e1896e45098055f4509a3b38525108dd7d6ac6e02d3f5c  Vito-Setup-2026.9.exe
cdca76fc98f3d7880165ad95a215d82a6f0011f570e2ed58c40d9a5785a08b14  Vito-2026.9.dmg
e188d28c1d570fabaa31f4498bcb8ce74238ae64e07e83ec534f5ee9d33aeac4  Vito-2026.9-x86_64.AppImage

Free, open-source, bring-your-own-key — no subscription. Setup guides at vito.talk.

⏰ Groq retires Vito's default model on 16 August 2026

If you use Groq for cleanup, update before that date. Groq decommissions llama-3.3-70b-versatile — the model Vito has shipped as its default — and llama-3.1-8b-instant alongside it. Left alone, your next dictation fails with a provider error.

The new default is openai/gpt-oss-120b, Groq's own mapping for the model being removed, so cleanup quality stays where it was. Existing configs migrate themselves on start: only those two exact model ids, and only against a groq.com endpoint — the same names on a local Ollama are still valid and are left alone.

The suggestion list is now the two models Groq designates production, plus two it flags preview. The preview ones carry a badge in the dropdown, because a model that can be withdrawn at short notice is exactly how this broke in the first place.

🐧 Linux: hotkeys and typing through XDG portals, and a Flatpak

Vito speaks the RemoteDesktop and GlobalShortcuts portals over D-Bus. On a Wayland desktop that means no ydotoold, no /dev/uinput udev rule and no root — you grant remote control once and it is remembered. A third backend, wtype, is preferred where it is available: it needs no daemon at all. ydotool remains the fallback for X11 and compositors without the portals, and an explicitly configured backend is always honoured.

Vito is now on Flathub-style hosting of its own: install from vinceecniv.github.io/Vito, signed with the repository key.

Which Linux download to pick:

Inside the sandbox a first dictation used to fail with ydotool not found in PATH — advice nobody could act on, since the sandbox has neither /dev/uinput nor a daemon, and the text then quietly went to the clipboard on every later attempt. Routes the sandbox cannot have are no longer offered, tried or named, and delivery falls back to the clipboard instead. A Flatpak on a desktop that cannot type now says so and points at the AppImage.

💸 OpenAI: current models, and a cost card that was wrong

The suggestions move to the GPT-5.6 family — luna for cleanup, terra for Assist, sol at the frontier. gpt-4o-mini stays in the list; it still works and is cheaper than luna.

The real fix is underneath: the cost card only ever knew gpt-4o-mini's price and showed it for every OpenAI model — off by 25× for anyone who picked the frontier model. Rates now follow the selected model.

🌍 Translations

The Linux and Flatpak strings, and the injection-backend labels that never reached the translator at all, are now in all 60 languages.

Install

• Windows — Vito-Setup-2026.8.1.exe (per-user, no admin).
• macOS — Vito-2026.8.1.dmg — drag Vito to Applications. First launch needs right-click → Open, then two permissions (Microphone, Accessibility).
• Linux — Vito-2026.8.1-x86_64.AppImage — chmod +x and run — or the Flatpak from vinceecniv.github.io/Vito.

Checksums (SHA-256)

1ac39c70b25aa96b8f34c49ba4504df8f5398af89c6589063667b2f325fb2834  Vito-Setup-2026.8.1.exe
739dbfb58f3aa702f2077693e3aeab529b9f78a6c7d623409828c6f691cbed34  Vito-2026.8.1.dmg
b9669fb5394442a72e558101ef6f0459df4c30b6cda507ac17b8ccc87bebba1c  Vito-2026.8.1-x86_64.AppImage

Free, open-source, bring-your-own-key — no subscription. Setup guides at vito.talk.

🍎 Vito now runs on macOS

Universal build (Apple Silicon and Intel), macOS 11 or newer. Vito lives in the menu bar, not the Dock.

Everything the other platforms do is there: global hotkeys with push-to-talk, paste/type/clipboard injection, notifications, launch at login, microphone level control and media ducking while you dictate.

The first launch takes a moment of ceremony. The build is signed, but with a self-issued certificate rather than an Apple Developer ID, and it is not notarised. Double-clicking a freshly downloaded copy shows "Vito cannot be opened": use right-click → Open, and if macOS still refuses — it does on recent versions — go to System Settings → Privacy & Security and press Open Anyway under the message about Vito. You only do this once.

Finder also shows a small ⌛ next to Vito's name. That is macOS reporting the app has not been through notarisation — not something still copying, and not something that approving the app clears. It is cosmetic; Vito runs perfectly well.

Then grant two permissions in System Settings → Privacy & Security:

🐛 Fixes for everyone

• The achievement chime never played, on any platform. The file's length fields were zero and the audio behind them was pure silence — playback reported success, so nothing ever surfaced it. Regenerated, and every embedded sound is now checked for actual signal.
• A failed AI cleanup is now audible. The "done" chime plays when recording stops, before cleanup runs, so a failed pass used to arrive as unpolished text with no signal at all. A distinct warning sound follows, and the notification says what happened.
• The welcome card, the privacy note and the cost note each listed only some of the providers Vito supports.

Install

• Windows — Vito-Setup-2026.8.exe (per-user, no admin).
• macOS — Vito-2026.8.dmg — drag Vito to Applications, then follow the first-launch steps above.
• Linux — Vito-2026.8-x86_64.AppImage — chmod +x and run.

Checksums (SHA-256)

6e6672cb71f1ee336ee3f30aa9a18cd7a11ab65b5757021a8d2b4b6d28814963  Vito-Setup-2026.8.exe
83320db116c16f9247444f8fda61d06062c56db6b06554135e761249c773c233  Vito-2026.8.dmg
a072c23f06a00560afcf7000bcb0dd248bd260f47f16637d1c1937e91723c2fc  Vito-2026.8-x86_64.AppImage

Free, open-source, bring-your-own-key — no subscription. Setup guides at vito.talk.

✨ Vito Assist — say "Vito, …" and it acts on your words

Start a dictation with "Vito, …" and a short instruction. It isn't typed out — it's carried out on your next dictation, or on the text already on your clipboard.

• Spoken commands — reshape your dictation ("translate to English", "summarise this", "turn into bullet points", "make it more formal"), or ask it something ("question" then ask, "reply to this"). The instruction runs through the AI-cleanup pass, the mic reopens on its own afterwards, and a unique sound confirms it's armed.
• Clipboard mode — "Vito, summarise the text on the clipboard" reads your clipboard and puts the result back on the clipboard, ready to paste — with a clear on-screen hint so it never feels like nothing happened.
• Live command banner with an animated Vito mark, in the app and the mini overlay.
• Its own (heavier) model, optional — borrow the AI-cleanup model, or point Vito Assist at a stronger one, since answering and transforming ask more than tidy-up does.

📊 Measured, and rewarded

• A dedicated "Vito Assist" status card counting how often you use it.
• A command achievement tier plus a hidden "Vito Wizard" for using all three modes — plain dictation, a spoken command and a clipboard command — with animated Noto medals.
• Honest cost split — Assist tokens are billed apart from cleanup, at their own model's rate. Both the AI-cleanup and Vito Assist settings now show editable token-rate fields, pre-filled from your provider (€0 on Groq's free tier), so you can match the exact price if you want.

⚙️ Settings & history

• A dedicated Vito Assist settings tab: how the flow works, examples, what it can't do, and the model configuration.
• The history and the timing bar now label an Assist dictation as "Vito Assist" and show the original command, instead of just "cleanup".

🌍 60 languages

• All the new interface strings are translated into all 60 languages.

🐛 Fixes

• The status-page layout editor no longer jumps around when you drag cards; the layout dialog is wider and centred over the content.

Install

• Windows — Vito-Setup-2026.7.3.exe (per-user, no admin).
• Linux — Vito-2026.7.3-x86_64.AppImage — chmod +x and run.

Free, open-source, bring-your-own-key — no subscription. Setup guides at vito.talk.

AI cleanup: OpenAI-compatible providers

The AI cleanup pass now works with any OpenAI-compatible chat-completions endpoint, alongside Anthropic:

• Groq — a free tier (rate- and daily-token-limited) that states it doesn't train on your API data. Together with AssemblyAI's free starting credit, you can dictate and clean up for free.
• OpenAI — use your own OpenAI API key.
• Local model — point it at Ollama, LM Studio or any local endpoint, so your transcript never leaves your machine.

New installations default to the Groq preset; existing configurations keep their current provider.

Downloads

• Linux: Vito-2026.7.2-x86_64.AppImage
• Windows: Vito-Setup-2026.7.2.exe

Verify each download against its .sha256 file.

Vito says Hello world!