macOS Voice Control vs Talon: an honest comparison

These two get compared constantly and usually unfairly. Apple's Voice Control gets dismissed as a toy by people who live in Talon, and Talon gets described as impossible by people who bounced off its setup in an evening. Neither is true, and the choice is more interesting than a winner.

What Apple gives you for free

macOS has shipped a genuinely full voice control system for years now. It is in System Settings, under Accessibility, in the Motor section. Switch it on and you can navigate, dictate, edit text, click things by name, and drive a numbered grid over anything that has no name to click.

What it does well is significant:

• It is already there. No download, no account, no configuration. The gap between deciding to try voice control and trying it is about ninety seconds.
• Clicking by name and number is excellent. Saying show numbers and then a number is a fast, reliable way to hit anything on screen, and it works in apps that were never designed with this in mind.
• The tutorial is good. Apple ships an interactive walkthrough inside the settings pane, which is more than most software in this space bothers with.
• It handles dictation and commands together without you having to think much about which mode you are in.

If you want to know whether talking to your computer suits you at all, this is the correct place to find out, and it costs you nothing but an afternoon.

Where people outgrow it

The limits are real, and they arrive in a predictable order.

Custom commands are awkward. You can add your own, but meaningful ones push you toward Apple's automation tooling, which is its own learning curve for what should be a small change. Wanting to say three words and have five things happen is a normal request that becomes a project.

Context is shallow. A command that should mean one thing in your code editor and something else in your browser is not something the built-in system is built around. Once you want that, you want a different tool.

There is no community. When something does not work, you are largely on your own with Apple's documentation. There is no shared set of commands that thousands of people have already refined.

It is comparatively slow to respond in a way that matters when you are issuing a lot of short commands in a row. This is subjective and hardware-dependent, but it comes up consistently from people who have used both.

What Talon is

Talon is a voice computing system built for people who intend to work this way permanently. Its speech recognition runs on your own machine, it works on macOS, Windows and Linux, and the public release is free.

Three things make it different in kind rather than degree:

Commands are declarative and scoped. You describe a spoken form and what it should do, and you can attach it to a particular application, a particular website, or a particular mode. This sounds like a small thing and it changes everything, because your vocabulary can be dense without being ambiguous.

The community command set is enormous and open. Most people run a shared, MIT-licensed set that covers thousands of spoken forms: text editing, window management, browsers, and a well-developed vocabulary for writing code out loud. It has been refined by people who use it every day because they have no alternative.

It was built for the hard case first. Talon's early community was largely programmers who had lost the use of their hands. Software designed around that constraint tends to be serious about reliability in a way that software designed around convenience is not.

Where Talon is hard

Almost all of Talon's difficulty is concentrated at the start, and almost none of it is about voice.

The conventional path is: download and install Talon, then separately find the community command set on GitHub, then clone or unzip it into a specific hidden folder inside your home directory, then restart Talon and hope. To change anything, you open .talon files or Python files in a text editor and learn a small syntax.

For a developer this is an unremarkable Tuesday. For everybody else it is a genuine wall, and it is the single most common reason people who need voice control never get it. We wrote about exactly where people give up and what to do instead.

There is a second, subtler difficulty. Once running, Talon knows thousands of commands and has no interface for showing you them. The knowledge lives in documentation, in a community wiki, and in a Slack channel. That is a big ask of somebody whose hands hurt and who just wants to answer their email.

Side by side

How to actually choose

Ignore which is more powerful and answer three questions.

Are you trying this, or committing to it? If you are still finding out whether voice control works for you, use Apple's. Every hour spent configuring something you might abandon is an hour wasted, and the built-in system will answer the question honestly.

Do you need to change things? The moment you catch yourself thinking "I wish I could say X and have Y happen", and that thought recurs weekly, you have outgrown the built-in system. That is the signal, and it is a much better one than any feature comparison.

Will this be how you work, or how you work sometimes? Full-time hands-free use rewards the investment in Talon quickly, because the friction you are removing is friction you meet hundreds of times a day. Occasional use may never justify it.

Common questions

The summary

Apple's Voice Control is the right first move for almost everyone and a permanently good answer for some. Talon is where you go when you have found the specific thing you cannot do, and it will not disappoint you once you are through the setup.

The gap between those two sentences is the problem worth solving, and it is why we built Starky.