Typing with RSI: Replacing the Keyboard with Your Voice

This post is practical guidance on shifting typing load to your voice when your hands need the break: what dictation can genuinely take over, what it can’t, and how to pace it. It is not a recovery story, and it is not medical advice. We make a dictation app. If typing hurts, a doctor or physical therapist is the right person for that conversation. We can only help with the tool side, which is what the rest of this post covers.

The framing that seems to help

When people with RSI write about going voice-first, the ones who stick with it rarely describe dictation as a cure or a keyboard replacement. They describe it as load shifting. Every sentence you speak is a sentence your hands didn’t type. On a good day that might be 20% of your output. On a bad day it might be most of it.

That framing matters because the people who try to go 100% voice on day one tend to bounce off. Dictation is genuinely good at some things and genuinely bad at others, and forcing it into the bad cases is frustrating enough that some people give up on the good cases too. Better to know the split going in.

Hold-to-talk beats always-listening, especially on bad days

Always-listening dictation sounds like the accessible option: no keys involved at all. In practice it tends to backfire. An open mic transcribes your sigh, your coworker’s question, the half-sentence you muttered while thinking. Then you’re doing cleanup edits, which means mousing and typing, which is the thing you were trying to avoid.

Hold-to-talk flips that. One key down, speak, key up, done. The mic is only hot while you decide it is. Holding a single key is a much smaller ask of your hands than typing the sentence would have been, and there’s nothing to clean up afterwards because nothing got transcribed that you didn’t say on purpose.

A foot pedal or a spare mouse button mapped to the dictation hotkey takes hands out of it entirely on flare days. We didn’t design Dictate around that, but macOS remapping tools make it work.

What dictation genuinely takes over

The workable split is consistent: prose goes to voice, structure stays on the keyboard in whatever small doses you can manage. Concretely:

Email. The biggest single win, because email is almost pure prose. Gmail on the web has no built-in voice typing at all, which surprises people. We wrote up how to dictate email in Gmail separately.

Documents and notes. First drafts especially. Speaking a rough draft and then lightly editing it involves far fewer keystrokes than typing the whole thing. A rhythm that holds up well: dictate the draft, take a break, then do one short editing pass by hand.

Chat. Slack and Messages replies are short, conversational, and constant through the day. Individually they’re nothing. Added up, they’re a lot of typing, and they dictate well.

AI prompts. The newest one, and for developers the one that changes the math. A good prompt to ChatGPT or Claude might be a paragraph long, it’s plain English, and you might write dozens a day. That’s the exact shape dictation is best at. Developers with RSI who write up their setups keep landing on the same point: the prompt-heavy style of AI-assisted coding is what made voice viable for their work at all. The AI writes the syntax, they speak the intent.

What it can’t take over

Precise code. Dictating useEffect(() => { into a general dictation app does not go well. Dictate is built for natural speech, not symbol-dense syntax. Comments, commit messages, docs, and prompts dictate fine. The code itself mostly doesn’t. If you need actual code by voice, see the Talon section below.

Spreadsheets. Cell navigation, formulas, selecting ranges: this is coordinate work, not prose. Voice text input helps with the occasional label or note and little else.

Fiddly edits. “Change the third word in that sentence” is where text-input dictation runs out. You can re-dictate the sentence, which is often quicker than it sounds, but surgical edits still want a cursor, whether that’s keyboard, mouse, or a voice control grammar.

No single app covers all of this. The realistic setup for most people is voice for prose, plus whatever amount of keyboard and mouse your situation allows, plus dedicated tools if you need to go further.

Pace your voice too

Move enough output to voice and a new budget shows up: your voice is also a part of your body, and it gets tired. Vocal fatigue at the end of a heavy dictation day is common enough to plan for. The common-sense mitigations: speak at a normal conversational volume (the mic is close, you don’t need to project), keep water around, and take the same kind of breaks you’d want from typing. If speaking itself becomes uncomfortable, that’s another thing for a professional, not for a blog post from a software company.

If you need more than text input

Dictate does one thing: speech in, text out, wherever your cursor is. For some people with severe RSI, text input is only part of the problem. The whole computer needs to be operable without hands. That’s a different job, and it needs different tools.

macOS Voice Control is Apple’s free accessibility feature. It does full hands-free control: open apps, click anything on screen by name or number, dictate, and correct text by voice. It’s more work to learn than plain dictation and the dictation part is less fluid, but nothing else built into the Mac comes close for full control. We wrote a comparison of dictation and Voice Control covering when each fits. Standard macOS Dictation is unavailable while Voice Control is on because Voice Control supplies its own dictation mode.

Talon is free software built for controlling a computer and writing actual code by voice, with an active community around it. The learning curve is real: you’re learning a spoken command grammar, not just talking. People who’ve put in the weeks report writing serious software entirely hands-free. If precise code by voice is your requirement, Talon is the right recommendation, not a general dictation app.

Starting without spending anything

Don’t buy anything yet, ours included. macOS has dictation built in. Turn it on, spend a few days speaking your email and chat instead of typing it, and see how the load-shifting feels. Our guide to dictating on a Mac covers setup and the built-in tool’s quirks.

The quirks are the reason paid apps exist. Apple’s dictation stops after about 30 seconds without speech, automatic punctuation can be uneven, and behavior changes from app to app. For a few sentences a day, it’s fine. If you’re moving real volume to voice because your hands need the break, the friction adds up precisely because you’re using it so much.

Where Dictate fits

Dictate is our app. Hold a hotkey, talk, release, and the text lands where your cursor was, in any Mac app. No pause timeout, punctuation handled, nothing to configure before it works. Transcription happens on our servers: audio goes over TLS, gets transcribed, and is discarded when the request completes. History keeps the text, never the recording. It’s $8 a month flat, one plan.

One warning so nobody gets burned: an unused subscription is refunded in full, but once you’ve transcribed anything, that month is non-refundable. So do the free experiment with Apple’s built-in dictation first. If a week of that convinces you voice belongs in your setup and the built-in friction is what’s getting in the way, that’s the problem the $8 solves.