Mac dictation does not know your words. Fix it in 2 minutes.

The native workaround, which does exist

You cannot add words to Dictation. You can add them to Voice Control, and the recogniser underneath is the same one, which is why this works at all. Apple documents it on its own support page.

1. Open System Settings.
2. Click Accessibility in the sidebar, then Voice Control.
3. Click Vocabulary.
4. Type the word or phrase exactly as you want it entered, then click Next.
5. Say it out loud, then click Save. If Dictation is off you can click Save Without Recording and add the pronunciation later.

Apple states a limit of 1000 terms. You can control click an entry to record or rerecord how it is said, and you can import a list from a text file with one term per line, or export what you have built.

What it does not solve

The import is the part worth noticing, because it tells you what this feature really is: a list you are expected to write in advance. That is fine for the twenty terms you already know you will need. It does nothing for the long tail, which is where the problem actually lives.

You do not know, before you start, that the recogniser will turn a particular library into two English words. You find out by watching it happen, at which point you are in the middle of writing something and are not going to open System Settings, click through to Accessibility, and type the word in. So you fix it by hand and it happens again tomorrow.

The other thing people reach for is text replacement under Keyboard. That is a different mechanism again: it swaps a string after the words have been recognised. It can rescue a term you can predict, and it cannot help with the hundred you cannot.

The faster fix

Zumbo comes at the same problem from both ends, and neither of them asks you to sit down and make a list.

In your note

Ship the Q3 roadmap before Friday.

Teach Zumbo this word

Heard ascue three Should beQ3

Replace here Replace everywhere and teach

Vocabulary packs. Twenty one of them, covering fields like developer tools, legal, medical, design, marketing, writing and research, and video and audio. You are asked what you dictate most during setup and the matching packs are switched on. Your field's language is there before you have typed anything, so product names, technical terms and commands come out spelled right the first time.

Teach. When a word does come out wrong, select it, type what you meant, press Teach. It is heard your way from then on. It takes about three seconds and happens at the exact moment you noticed the problem, which is the only moment you were ever going to fix it.

The result is a vocabulary that is a byproduct of working rather than a project. After a week it matches what you actually write, including the internal names that are not in anybody's model.

Which one should you use

If you have a short, known list of terms and you are happy with the built in dictation otherwise, the Voice Control vocabulary is free and it works. Use it. Set aside ten minutes, type your twenty terms, record the awkward ones.

If the problem is that you keep discovering new wrong words in the middle of your work, a list is the wrong shape of solution and no amount of filling it in will catch up. That is the case Zumbo is built for.

Related: the 21 vocabulary packs, the Apple Dictation alternative that does not time out, and dictation for developers.