The best free dictation software

The free dictation software that actually exists

Here's the field, with the one thing each tool's marketing page would rather you skim past. The "genuinely free?" column is the honest one.

A few of these need unpacking, because the catch is the whole story.

The built-ins are real, and they have edges

Your operating system probably ships with dictation already. On Windows 11 you press Windows key + H and a small voice bar appears; you talk, the words land wherever your cursor is. It's free, it's built in, and for a quick reply it's genuinely fine. The catch is that Windows Voice Typing uses online speech recognition powered by Azure, so it needs an internet connection, and your audio goes to Microsoft's servers to come back as text.

On a Mac, Apple Dictation does the same job from the keyboard. It can run on-device after you download the speech model, which means it works offline and your voice can stay on the machine. It supports dozens of languages and regional variants; Apple lists 54 online, 43 available offline. The honest gaps: it's Mac-only, there's no custom vocabulary, and there's no AI cleanup pass to fix the rambling and the "um"s.

Both are good. Neither asks for a cent. The reason people still go looking for something else is usually one of three things. They're on the wrong OS. They want their audio to stay offline without exceptions. Or they've outgrown short notes and want the transcript to come out already cleaned up.

The "free" tier that isn't quite free

This is where the word does the heavy lifting. Several well-made dictation apps advertise a free tier, and they mean it, right up to the point that matters.

superwhisper is a strong, mature app that runs on Mac, Windows, and iOS, and it genuinely works offline. But the offline part, the local voice models, is Pro-only. The free tier gives you limited cloud models, which is the opposite of why most people want a local tool. (If you're weighing it specifically, I wrote a longer superwhisper comparison that gives it full credit.)

Willow Voice is free with no credit card, which is refreshing, until you read that the free plan is 2,000 words per week. That's roughly four decent emails before you hit the wall, and there's something uniquely deflating about a word counter stopping you mid-sentence on a Thursday.

Voice In is free and simple and does exactly what it says. It's a Chrome extension, which means it only works inside the browser and needs an internet connection. The moment you want to dictate into a desktop app, your editor, your email client, anything that isn't a Chrome tab, it can't follow you there.

None of these are bad tools. They're just free in a way that has an asterisk, and the asterisk is the part you'll meet on day three, not day one.

Open-source free is real free, if you're willing to build it

If "free" to you means "no company can ever charge me", open source is the honest end of the spectrum. OpenAI's own Whisper is released under the MIT license, code and weights, and whisper.cpp is a fast C/C++ port under the same license. VoiceInk is a Mac app under GPL-3.0, with the source on GitHub for anyone to read or build.

The catch is the work. OpenAI Whisper installs through pip, needs the ffmpeg tool alongside it, and runs from the command line to transcribe audio files. It's not a hotkey that types into your apps; it's a tool that turns a recording into a transcript. VoiceInk is a polished app, but it's macOS 14.4 and Apple Silicon only. For a developer who enjoys the setup, this is paradise. For everyone else, "free" stops feeling free around the second error message.

What free looks like when nothing runs out

I built Whisper by Remskill on the side starting in 2024, and the one rule I didn't want to break was the bait-and-switch. So the entire local pipeline is free for any signed-in user, and there's no card at signup.

Free here means the actual work, not a teaser. You get the Whisper transcription models and Parakeet, local AI cleanup through Ollama, your dictation history, custom hotwords, presets, and a hotkey you can remap. No weekly word cap. No "local models are Pro-only." It runs on Windows and Mac. You press a hotkey (Ctrl+Space on Windows by default, remappable), talk, and the text lands where your cursor is.

The paid part is honest about being paid. Whisper Pro adds the cloud surface: OpenAI cloud transcription, cloud AI cleanup, and live web search. You bring your own OpenAI key, so we take no cut. The numbers live on the pricing page, not buried in an app you already installed. If you only ever use the free local side, that's a complete product, not a countdown.

When a free built-in is all you need

Here's the part the makers of paid dictation apps don't print on their homepage. For a lot of dictation, the free thing already on your computer is the right answer, and you should not install anything.

If you're firing off a 30-word text, a one-line Slack reply, or a quick note to yourself, open Windows Voice Typing with Win+H or hit Apple Dictation from the keyboard and move on. It's free, it's instant, and there's no account. Don't reach for a dedicated app, free or paid, to do something your OS does for nothing.

The built-ins start to hurt around the longer stuff: a paragraph that needs to come out clean, a transcript in a language your OS doesn't cover well, audio you'd rather keep offline with no exceptions, or the same correction you make every single time because there's no custom vocabulary. That's the threshold where a dedicated tool earns its place. Below it, save yourself the download.

I think about this every time someone asks what to use. The honest answer usually starts with "what your computer already has", and only then moves on to anything you'd install. My own mother got there before I did; she'd been dictating shopping lists with the built-in for a year before she asked me what my app was for.

How far free has come

Sometime in the late 1990s, a relative had Dragon NaturallySpeaking on a Windows 98 machine with 64 MB of RAM. Setting it up meant reading a list of words aloud for 45 minutes so the software could "calibrate". After all that, dictation worked at maybe 70% accuracy with a four-second delay per sentence. It took fifteen minutes to dictate one paragraph of a holiday letter. The headset got thrown across the room. The headset survived; the experiment did not.

I tell that story because the bar for "free dictation software" today would have been science fiction then. No training. No four-second wait. Models that handle accents and dozens of languages out of the box, running on the laptop you already own, for nothing. The thing the headset-throwers wanted is now the floor, not the ceiling, and most of it costs zero.

If you only remember one thing

Try the free built-in first. Win+H on Windows or Apple Dictation on Mac covers more of your dictation than you'd expect, and it's already paid for. When you outgrow short notes (longer text, offline-only, cleaner output, more languages) pick a tool whose free tier doesn't end mid-sentence. That's the whole test: not "is it free", but "does the free part run out". For the local side of Whisper, it doesn't. If you want the wider field of dictation tools, the voice typing software roundup walks through it, and the Windows voice-to-text guide covers the built-in step by step.