A Plain-English Glossary of Dictation Terms
Dictation on a Mac is simple to use but wrapped in confusing words: speech to text, on-device, VAD, WPM, post-processing. This plain-English glossary explains the terms you actually meet, so you can pick a tool and start talking without a computer-science degree.
Key takeaways
- Dictation and speech to text mean nearly the same thing from two angles: one is the act, one is the engine.
- On-device is the term that matters most for privacy: it means your audio never leaves your Mac.
- VAD, WPM and models sound technical but describe simple, everyday parts of how dictation works.
- AI cleanup turns messy spoken words into polished, punctuated text automatically.
Why the words matter
When you shop for Mac dictation, the marketing throws around terms like "on-device Whisper" and "post-processing" as if everyone already knows them. Most people do not, and that makes it hard to compare tools fairly. Once you can tell the difference between where a model runs and how it cleans up your text, the choices get a lot clearer. If you want a ranked overview after this glossary, our guide to the best dictation software for Mac in 2026 puts these ideas into practice.
This page is a reference. Skim it, bookmark it, or read it top to bottom. Each term below is defined in one or two sentences, in language you could repeat to a friend.
The core dictation terms, defined
These are the words you will see most often on a Mac. Start here.
- Dictation
- The act of speaking out loud so a computer writes the words for you. On a Mac you trigger it with a shortcut, talk, and the text appears where your cursor is.
- Speech to text (STT)
- The underlying technology that converts spoken audio into written words. Dictation is the experience, speech to text is the engine that makes it happen.
- Voice to text
- A friendlier name for the same idea as speech to text: your voice goes in, text comes out. The terms are used interchangeably in everyday writing.
- Transcription
- The written output of speech to text. When a dictation app turns your recording into a paragraph, that paragraph is the transcription.
- On-device (local)
- The speech recognition model runs on your Mac's own hardware instead of a remote server. Your audio is processed locally and never uploaded, which is what keeps it private and able to work offline.
- Cloud dictation
- The opposite of on-device: your audio is sent over the internet to a company's servers to be transcribed. It can be accurate but means your voice leaves your machine.
The single most important distinction above is on-device versus cloud. If a tool transcribes on your Mac, like BlaBlaType does, your audio and transcripts never leave the device. If it uses the cloud, they do. For anything sensitive, that is the line that matters.
The technical terms, without the jargon
These words sound complicated but describe simple parts of the pipeline. You do not need to configure any of them to dictate, but knowing them helps you read reviews and settings screens.
- Model
- The trained software that recognizes speech. Popular on-device families include Whisper and Parakeet. A bigger model can be more accurate but uses more memory.
- Whisper
- A widely used open speech recognition model that runs well on a Mac without the cloud. Many local dictation apps, including BlaBlaType, use it or a variant.
- Parakeet
- Another on-device speech recognition model known for being fast and light, often used as a quicker alternative to larger Whisper models.
- VAD (Voice Activity Detection)
- The part of a dictation app that tells speech apart from silence, so it only transcribes when you are actually talking. Good VAD trims pauses and background noise without clipping your words.
- WPM (words per minute)
- A measure of how fast text is produced. Comfortable speaking sits around 120 to 150 WPM while typing is closer to 40, which is why most people speak around three to four times faster than they type.
- Post-processing (AI cleanup)
- A step that rewrites your raw transcription into polished text: removing filler words, fixing punctuation and grammar, and adapting tone. In BlaBlaType this runs on-device using Apple Intelligence.
- Custom dictionary
- A short list of names, brands and jargon you add so the app spells them correctly every time, instead of guessing at unusual words.
WPM is worth pausing on, because it explains why people switch to voice in the first place. If you want the background on how the number is measured, the concept of words per minute has a long history in typing and stenography.
How the terms fit together in one session
Here is the whole vocabulary in motion. This is exactly what happens between pressing a shortcut and seeing finished text, with the glossary words in bold.
You start dictation
You press a shortcut and speak. This is the dictation step: your voice becomes the input for everything that follows.
VAD listens for speech
Voice Activity Detection separates your words from silence and background noise, so only real speech moves forward.
The model transcribes on-device
An on-device model like Whisper or Parakeet turns audio into a raw transcription on your Mac, with nothing uploaded.
AI cleanup polishes the text
Post-processing removes filler, fixes punctuation and grammar, then drops the finished text wherever your cursor is.
Put together, those four steps are why dictating an email feels effortless once you get used to it. If email is your main use case, we walk through it in detail in how to dictate emails on Mac.
Terms for pricing and access
A few words come up when you compare what these tools cost. They are worth knowing before you commit.
| Term | What it means | Why it matters |
|---|---|---|
| Free trial | A limited-time way to test a paid app | A no-card trial lets you try before paying anything |
| Per-minute billing | You pay for each minute of audio transcribed | Common with cloud tools, and it adds up fast |
| Flat plan | One recurring or one-time price | Predictable cost regardless of how much you dictate |
| Translate as you speak | Speaking one language and getting text in another | Useful for multilingual work across 90+ languages |
On price, on-device tools have a structural advantage: because there is no server transcribing your audio, there are no per-minute cloud costs. If cost is your deciding factor, see the cheapest way to dictate on a Mac in 2026, and compare full plans on our pricing page.
Put the glossary to work
On-device speech to text, VAD, AI cleanup and 90+ languages, all running privately on your Mac. No card needed for the trial.
Download for macOSOne last note on the built-in option. Apple's own Mac dictation feature is free and uses many of these same ideas, so it is a fine way to learn the vocabulary hands-on before deciding whether you want AI cleanup and a fully on-device workflow.
Frequently asked questions
What is the difference between dictation and speech to text?
They describe the same thing from different angles. Dictation is the act of speaking so a computer types for you. Speech to text is the underlying technology that converts spoken audio into written words. On a Mac, you use dictation, and speech to text is what makes it work.
What does on-device dictation mean?
On-device dictation means the speech recognition model runs on your Mac's own hardware instead of a remote server. Your audio is transcribed locally and never uploaded, which is why on-device tools like BlaBlaType keep every word private and can work offline.
What is a good words-per-minute rate for dictation?
Comfortable speaking runs around 120 to 150 words per minute, while typing usually lands near 40. Most people speak around three to four times faster than they type, which is why dictation feels quick once the vocabulary and punctuation are cleaned up.
What is VAD in a dictation app?
VAD stands for Voice Activity Detection. It is the part of a dictation app that tells the difference between speech and silence, so the app only transcribes when you are actually talking. Good VAD trims background noise and pauses without cutting off your words.
What is AI cleanup in dictation?
AI cleanup is a step that rewrites raw transcribed speech into polished text. It removes filler words, fixes punctuation and grammar, and can adapt tone. In BlaBlaType this runs on-device using Apple Intelligence, so your text is cleaned up without leaving your Mac.