The modern dictation app

Press a hotkey, talk, and the words land at your cursor

Here is the mechanic that defines a desktop dictation app. You press a key, Ctrl+Space on Windows by default, a remappable push-to-talk shortcut. You speak a sentence. You release the key. A second or so later, the text shows up exactly where your cursor was blinking.

The "at your cursor, in any app" part is what separates a real dictation app from a transcription website. A transcription tool takes a recorded file and gives you back a wall of text you then have to copy somewhere. A dictation app skips the file entirely. The thing you are typing into is the thing the words appear in. It is the difference between dictating a letter and transcribing a recording of yourself reading one.

The speed matters more than people expect. Typing runs around 40 words a minute for most of us. Speaking runs around 145. That is roughly three and a half times faster, which is why a tool you press for one second can save you a real chunk of an afternoon. The email backlog, the meeting notes, the twelve cold-email variants before lunch.

Your computer already has one, and sometimes that is enough

Before you install anything, know that you almost certainly already own a dictation tool. Both major operating systems ship one for free.

On Windows 11, you press the Windows key + H and a voice typing bar appears. It works, with one catch: it uses online speech recognition powered by Azure, so it needs an internet connection to do anything. No connection, no dictation. It can insert punctuation for you if you turn that on.

On the Mac, Apple Dictation lives in System Settings under Keyboard, started with the microphone key or a shortcut you pick. Apple's version can run entirely on-device once you download the speech model, which means it works offline and your voice does not leave the machine. It covers dozens of languages. Apple lists 54 online and 43 available offline on its feature-availability page. What it does not do is custom vocabulary, AI cleanup of your phrasing, or letting you pick which transcription model runs.

The boring truth is that for short, casual dictation, a 30-word text, a quick search, a one-line reply, the built-in tool is genuinely fine. It is free, it is already there, and you do not need a fourth icon in your menu bar to send your mum a message. If that is all you do, close this tab and go press Win+H or your Mac's dictation key. I mean it.

Where a dedicated dictation app earns its place

So when does the built-in stop being enough? Three things, usually. The first is offline that is actually private. Windows voice typing needs the cloud. A dedicated app can run the whole thing on your own CPU, so the email to your kid's school or the legal brief you are drafting never travels anywhere. The second is accuracy and control: picking a bigger model, adding custom words it keeps mishearing, having your phrasing cleaned up automatically. The third is the part the built-ins quietly skip, which is working the same way in every app, every time, without surprises.

This is the slot Whisper by Remskill sits in. It is a desktop app for Windows and macOS on Apple Silicon. The entire local pipeline is free for anyone who signs in, no card at signup. That includes the local transcription engines, AI cleanup, history, custom words, and registering up to three of your own devices.

For the engine, you are not handed one default and told to like it. You pick. Local Parakeet is the fastest option: about 600 MB, 5 to 10 times faster than Whisper on a CPU, covering English plus 24 other European languages. Local Whisper is slower but supports 99 languages, can translate speech to English, and gives you custom vocabulary and finer control; the largest model is about 3 GB. And if you want OpenAI's latest cloud models and live web answers, there is an optional Pro cloud surface where you bring your own API key and Remskill takes no cut. Three paths, you choose based on whether you care most about speed, languages, or the newest cloud quality.

The honest opinion: most "AI dictation" apps are a markup, not a model

Here is my one strong take, and I will back it with a number. A lot of the dictation apps launched in the past two years are the same thing under the hood: a speech model you could run yourself, a clean UI on top, and a monthly invoice that has very little to do with the cost of the compute. The going rate for that arrangement lands around thirty dollars a month for the pricier ones.

The pitch deck says "powered by AI." So is my mortgage application. The phrase has stopped meaning anything. What you are actually paying for, in a cloud-only app, is the convenience of someone else running a model, plus the privacy cost of your audio leaving your laptop to do it. If the work is sensitive, that is not a convenience, it is a liability.

That is the whole reason I am stubborn about free local being the floor, not the upsell. Your laptop has a microphone and a CPU. For one paragraph of dictation, it does not need a server in the loop.

A quick, honest tour of the other dictation apps

You should know the landscape before you pick. Here are the ones worth naming, one line each, credit where it is due.

I left a few out. The point is not to memorize a list. It is that "dictation app" is a crowded word covering free built-ins, open-source projects, polished subscriptions, and enterprise voice-command suites. Pick by what you actually do, not by whose homepage you saw first.

Why "just press one key" is harder than it sounds

A short story, because it is the reason I care about the boring details. I gave the app to my younger daughter, who is seven, on a Saturday. One demo: press, talk, release, paste. She wrote a 90-word email to her grandmother about a lost tooth, the tooth fairy's exchange rate, and dance class, without asking a single question.

Two days later she came back: "the hotkey doesn't work in my drawing app." The average person does not know what a hotkey conflict is. They just know it stopped working. That night I shipped the customizable-hotkey settings. A dictation app lives or dies on the small stuff: the conflict you did not predict, the phantom key release on Windows that once made my handler fire six times per press and cost me several days and a 300ms debounce to fix. I have a master's degree. The debounce still humbled me.

When you don't need a dictation app

This is the part most articles skip, so here it is plainly. You do not need a dedicated dictation app if any of these is true.

• You only dictate short, casual things. Texts, searches, the odd one-line reply. The built-in tool handles that for free.
• You are always online and never touch anything sensitive. Then the cloud built-ins, or a cloud-only paid app, are fine; you are not gaining much from local.
• You only ever dictate inside one place, like a single browser tab. A browser extension covers that without a system-wide app. Voice In, for instance, is a Chrome extension that works only inside the browser.
• You want hands-free control of your whole computer, not just text entry. That is voice-command territory, and Dragon does it better than any dictation-first tool.

A dedicated app starts earning its keep at the longer, more frequent, more private, more multilingual end, roughly past the point where you are dictating real paragraphs, every day, and would rather they not leave your machine. Below that line, save your money. For a fuller comparison of the free options, there is a guide to free dictation software, a broader look at voice typing software, and if you came here from one specific app, the superwhisper alternative breakdown.

If you only remember one thing

A dictation app is just "press a key, talk, the words appear where you're typing." The built-in one on your computer already does the easy version for free. A dedicated app is worth it when you need offline privacy, real multilingual accuracy, or the same behavior in every app, and worth skipping when you don't.

Whisper by Remskill keeps the whole local pipeline free for exactly that reason: you should be able to find out where the line is for you without paying first. Pricing for the optional cloud Pro tier lives on the pricing page. My seven-year-old found the line in one Saturday. You will probably be faster.