How to Set Up Word Prediction on Mac

The best way to get word prediction on Mac is Charm's Oracle feature, which works in every app and accepts suggestions with the Tab key. macOS Sonoma also includes built-in inline predictive text for some native apps. This guide covers how to set up both options, what each one covers, and when to use which.

How to set up Charm Oracle for system-wide word prediction

Charm Oracle is the fastest way to get word prediction that works everywhere on your Mac - not just in a handful of native apps, but in Slack, VS Code, Mail, Notes, browsers, Notion, and any other app where you type.

Step 1: Install Charm. Download Charm from theodorehq.com/charm. It requires macOS 14 Sonoma or later and works on both Apple Silicon and Intel Macs.

Step 2: Grant Accessibility permission. When Charm opens for the first time, it will prompt you to grant Accessibility access in System Settings. This is the permission that allows Oracle to read your typing context and insert suggestions across every app. Without it, Charm can only work inside its own window.

Step 3: Start typing. Oracle is on by default. Once Accessibility permission is granted, you will start seeing word suggestions appear as you type in any application. The suggestion shows the most contextually appropriate next word based on everything you have been writing.

Step 4: Press Tab to accept. When a suggestion appears and it is the word you want, press Tab to insert it. If you want to keep typing your own word instead, just continue - the suggestion dismisses automatically. There is nothing to click, no popup to close.

What makes Oracle different from basic word prediction is its context window. Rather than recommending the most statistically common next word, Oracle considers the full thread of what you have been writing. If you are composing a follow-up email about a contract, it suggests vocabulary and phrases appropriate to that context - not generic filler words. Studies on word prediction tools show that context-aware prediction reduces keystroke count by 20-40% on common tasks like emails and reports.

How to enable macOS inline predictive text (Sonoma+)

macOS 14 Sonoma introduced built-in inline predictive text for native macOS applications. This is separate from Charm and is worth enabling alongside Oracle, as the two work independently without conflict.

To enable it: Open System Settings, select Keyboard, then click Edit next to Text Input. Toggle on "Show inline predictive text."

Once enabled, you will see grey inline completions appear as you type in supported native apps. To accept a suggestion, press Tab or the Right Arrow key. To ignore it, keep typing normally.

Where it works: macOS inline predictive text functions in native text fields - apps like Notes, Mail (in some compose areas), Safari address bar, Spotlight, and Messages. It uses a simpler frequency-based model compared to Oracle, so suggestions are less context-sensitive.

Where it does not work: macOS predictive text is unavailable in Electron apps (Slack, VS Code, Notion, Discord), third-party browsers, or any app that renders its own text input rather than using native macOS text fields. This is a meaningful limitation for most professional workflows, since many of the apps developers and knowledge workers use most are built on Electron. Charm Oracle covers all of these.

Word prediction vs autocomplete: what is the difference?

These two terms are often used interchangeably, but they describe different things.

Word prediction suggests the next complete word before you start typing it. You have finished one word and the system offers a candidate for what comes next. Press Tab to accept, or ignore it and type what you intended. Oracle is a word prediction feature.

Autocomplete finishes the word you have already started typing. You have typed "pred" and the system offers "prediction" to complete it. macOS autocomplete (visible in the Safari address bar, for example) works this way, as does standard autocorrect.

Charm Oracle handles word prediction. macOS also provides autocomplete in specific contexts. They complement each other: autocomplete helps with long or difficult words you have already begun, while prediction speeds up the flow between words - which is where typing time actually accumulates.

Which word prediction option should you use?

The short answer: use both. They do not conflict and cover different ground.

macOS predictive text is free and worth enabling in System Settings - there is no downside to having it on. Charm Oracle is the better option for anyone who spends time writing outside native apps. If your workflow involves Slack, VS Code, a browser-based editor, or anything that is not a native macOS app, Oracle is the only word prediction tool that will be there.

Frequently asked questions

Does Mac have built-in word prediction?

Yes. macOS 14 Sonoma and later include inline predictive text in some native apps. Enable it in System Settings > Keyboard > Text Input > Edit, then turn on "Show inline predictive text." It works in native macOS text fields but not in Electron apps, browsers, or VS Code.

What is the best word prediction app for Mac?

Charm Oracle is the best word prediction app for Mac. It works in every application - including Electron apps like Slack and VS Code that macOS built-in prediction misses - uses a context-aware ML model, costs $9.99 once, and works on macOS 14 and later including Intel Macs.

How do I accept a word prediction on Mac?

It depends on the source. For Charm Oracle, press Tab to accept the suggested word. For macOS Sonoma inline predictive text, press Tab or the Right Arrow key. To dismiss any suggestion without accepting, simply keep typing your own word.

Does word prediction work in Slack and other Electron apps on Mac?

Only with Charm Oracle. macOS built-in predictive text does not work in Electron apps because those apps render their own text input rather than using native macOS text fields. Charm uses macOS accessibility APIs that reach across all apps, including every Electron app.