Ghost text is the UI pattern where a suggestion appears inline, right after your cursor, rendered in a lighter grey than the text you've actually typed. Keep typing and it's discarded silently. Press Tab and it becomes real text. It's called "ghost" text because it isn't committed to the document — it's a preview that vanishes the moment you ignore it.

If you've used GitHub Copilot in an editor, you've seen it. If you've typed a URL into a browser address bar and watched the rest of a previously visited site appear in grey, you've seen a simpler version of the same idea. The pattern is old — command-line shells have offered grey inline history completion for decades — but its use for full-sentence natural-language suggestions is newer, and Windows has been slow to adopt it outside of narrow contexts.

Where Ghost Text Already Exists

Code editors

VS Code, JetBrains IDEs, and similar tools show inline completions for code — from simple bracket-matching to AI-generated multi-line suggestions.

Browser address bars

Chrome, Edge, and Firefox complete URLs from history in grey text as you type, accepted with the right arrow key or Tab.

Shells and terminals

Fish, Zsh with autosuggestions, and PowerShell's PSReadLine module all show grey inline completions from command history.

Word, narrowly

Word occasionally offers grey inline completions for dates and recognized phrases, accepted with Enter — but only inside Word documents.

Why It's Rare Outside Those Apps

Ghost text looks simple, but it requires an app to do two things well: render an overlay that sits visually inside a text field without becoming part of the actual text, and generate a plausible suggestion fast enough that it doesn't lag behind typing. Code editors and browsers have both pieces because completion is core to what they do and their vendors invested in it directly.

Most Windows apps — messengers, mail clients, general text editors, chat tools — were never built with a completion engine in mind, and adding one is a per-app decision made by that app's own developer. There's no shared Windows API that hands every text box a ghost-text layer for free. The result is that the pattern spread wherever a vendor cared enough to build it, and stayed absent everywhere else.

The gap in one sentence: ghost text isn't hard to build once — it's hard to build into every app individually, which is why it clusters in code editors and browsers instead of spreading system-wide.

Bringing Ghost Text to the Rest of Windows

Typeahead is built to close that gap without waiting for every app vendor to add the feature themselves. It runs as a system-tray utility and renders an inline ghost-text suggestion at your caret in almost any Windows desktop app — WhatsApp, Telegram, Slack, Chrome, Edge, Firefox, Outlook, Thunderbird, Word, Notepad, Notepad++, and VS Code's chat and text panes, with code buffers off by default so it doesn't compete with a dedicated code completion tool.

Suggestions are generated by a local LLM (Qwen 3 via llama.cpp) running fully offline, or by custom snippet autocorrect and dictionary-based autocomplete when no model is loaded yet. Press Tab — the default key, and configurable — to accept, exactly like the pattern you already know from a code editor or browser address bar. Nothing typed leaves the device: no account, no text telemetry, and password fields and financial apps are excluded automatically.

Frequently Asked Questions

What exactly is "ghost text"?
It's an inline text suggestion rendered in grey right after your cursor. It isn't part of your actual text until you explicitly accept it, usually by pressing Tab; otherwise it disappears the next time you type or move the cursor.
Is ghost text the same as autocomplete?
Ghost text is one way to present autocomplete — inline and grey, versus a dropdown list of options. Both aim to save keystrokes; ghost text is generally faster to accept because there's nothing to select from, just Tab.
Why does VS Code have ghost text but Slack doesn't?
Code editors built completion engines into their core because it's central to writing code. Chat apps and messengers weren't designed around text prediction, so the feature only appears where a developer chose to add it — which most haven't.
Does Typeahead work in code editors too?
In VS Code, Typeahead applies to chat and text panes, not code buffers — code buffers are off by default so it doesn't conflict with dedicated code-completion tools you may already use.
Does ghost text from Typeahead require the internet?
No. Suggestions come from a local LLM (Qwen 3 via llama.cpp) that runs fully offline once downloaded, or from snippets and dictionary autocomplete that work even before a model is installed.

Get Typeahead

One-time purchase, no subscription. Windows 10 / Windows 11. 7-day free trial.

 Get it on Microsoft Store