Best Dictation Software for Windows: Free and Paid

For most people, the best dictation software on Windows is the free dictation already in the PC. Win+H voice typing works in any text box, but it needs an internet connection. Voice access in Windows 11 can control the PC and write text without internet once it is set up. Pay for an app when you need one of four things: processing that stays on the PC, accuracy on long documents and specialist words, AI cleanup of your sentences, or spoken control of every program. Dragon Professional v16 is the long-standing choice for command-heavy work. Wispr Flow and Aqua Voice are cloud apps for everyday writing. Handy is a free open-source app that transcribes on your own computer.

This guide covers the free tools first, then the paid and open-source groups, then the check that matters most: how the app scores on your own voice and vocabulary. For Windows setup steps, see our Windows voice typing guide and the Voice Access guide. For the comparison across all platforms, see the best dictation software guide. Yap, which we make, is an on-device dictation app for Mac and Windows, so we have a stake here. We have limited this guide to what the sources say.

The free built-in tools are enough for short dictation

Windows includes two free dictation tools. Voice typing is the quick one. On Windows 11 you turn it on by pressing the Windows key and H on a hardware keyboard ¹. It needs a working microphone, an internet connection and the cursor in a text box ². For short messages, notes and emails, that is often all you need.

Voice access is the fuller tool. Microsoft says it is available in Windows 11, version 22H2 or later ³. Its first-time setup needs an internet connection ⁴. After that, Microsoft says it lets you control your PC and write text using only your voice, without an internet connection ⁵. It is the better choice for hands-free work, because it can control your PC and edit text, not only type it.

The two tools differ in ways that matter for choosing. Voice typing sends speech to Microsoft's online service, so it is the wrong choice for a document that must stay on the PC during dictation. Voice access works offline after setup, but it needs Windows 11 22H2 or later, so it is not an option on older machines. Check which version your PC runs before you plan around either tool.

The built-in tools are enough when you write short texts, your room is quiet and you can accept some corrections. They are a weaker fit for long documents, unusual vocabulary and any text that must not leave the machine.

Windows dictation software falls into five groups: built-in, Dragon, cloud AI apps, on-device apps and open-source tools

The apps on Windows differ mainly by where they process your voice and how they are paid for. The table groups them, with the examples this guide checks.

The grouping matters more than the brand names. A cloud app and an on-device app can both be accurate. They differ in what happens to your audio, what the app costs to run, and whether it still works without a connection.

Where your audio is processed matters more than the feature list

Windows separates speech recognition into two modes. Microsoft's privacy article says Windows 11 has device-based recognition, which "Processes your voice locally on your device. No voice data is sent to Microsoft." ⁶ Online recognition works differently. Microsoft says voice data is sent to Microsoft only to provide the service and create text transcriptions, and that it does not store voice recordings without your permission ⁷.

The article does not say which dictation features use the device-based mode, so do not assume that Win+H works that way. Microsoft describes voice typing as online, so read the setting under Privacy & security in Windows Settings before you dictate anything sensitive. If you cannot confirm where a feature sends audio, treat it as cloud dictation.

The cloud apps are more explicit. Wispr Flow says it deletes your audio as soon as it has been transcribed, unless you have turned on its Dictation cloud storage setting ⁸. Aqua Voice says that with Privacy Mode turned off, it may store transcript data on its servers to improve the product ⁹. Both are policies you can change, so check the setting you are using, not just the default.

The on-device tools make the opposite claim. Handy, a free open-source app, says in its README that transcription happens on your own computer without sending any information to the cloud ¹⁰. Superwhisper says on its home page that it works offline ¹¹.

Read the privacy page of any app for the mode you actually use. Many apps send audio to the cloud only when you choose a cloud model or a cloud feature, so a local-by-default app can still send audio if you switch models.

Dragon Professional remains the long-standing choice for command-heavy professional dictation

Dragon Professional v16 is the current Windows release of Nuance's product. Its vendor page says it "is optimized for Windows 11 and backwards-compatible to Windows 10" ¹². Dragon's main strength is control: the computer and its programs respond to spoken commands, which is why it suits people who want to work by voice for hours at a time rather than dictate short passages.

The vendor page also claims that employees can "dictate documents 3 times faster than typing with up to 99% recognition accuracy" ¹³. That is the company's own figure. We did not find an independent test that measures the same thing on the same audio, so read it as a claim to test, not as a result. Your own test in the last section will show whether the claim holds for your voice and vocabulary.

The main barrier is price. A US reseller lists a one-user Dragon Professional v16 license at $685.99 ¹⁴. The Nuance product page we read shows no price and points to a contact page for licensing, so the price you pay will depend on the reseller and the plan.

Dragon also comes as a subscription. One reseller lists Dragon Professional Anywhere at $660 per year ¹⁵. A subscription spreads the cost over time, and a license is a one-time purchase, so compare the first two years, not the sticker price alone.

We did not find any statement in Nuance's product page that Dragon Professional 16 runs natively on Windows on Arm. If you use a Windows on Arm device, confirm this with Nuance before you buy. Our Windows on Arm dictation guide covers the wider question.

Cloud AI dictation apps polish your text as you speak and send it to a server to do so

Cloud AI apps are built for everyday writing. They tidy sentences while you speak, so you get a draft that needs less editing. The trade-off is that your audio travels to a server, and the free tiers have limits. Start with the limits, because they decide whether a free plan is enough for a day of work.

Wispr Flow's free plan allows 2,000 dictated words per week on desktop ¹⁶. Heavy daily use can use up that weekly allowance early in the week. Its Pro plan is listed at $12 per user per month ¹⁷, which comes to $144 over a year when billed annually ¹⁸. The same pricing page lists $15 per user per month ¹⁹, which is $180 over a year at the monthly list price ²⁰, so check the billing period before you compare prices.

Aqua Voice is available for Mac, Windows, iOS and Android through Google Play ²¹. Its Pro plan costs $8 per month when billed annually ²², which is $96 for a year ²³. It costs more on monthly billing, at $10 per month ²⁴, so decide the billing period before you compare prices.

When you compare cloud apps, check four things: the free-tier limit against your real daily use, the billing period, whether the app stores transcripts by default, and whether you can switch off any setting that uses your audio for model training. Those four points matter more than a small difference in monthly price.

On-device dictation apps keep audio on the PC and work offline

On-device apps keep the recognition on your machine, which matters for confidential notes and for working without a connection. Superwhisper's home page lists its dictation for Mac, Windows, iOS and Android, but its meeting transcription only on Mac ²⁵. The same page says the app works offline ¹¹. Check the feature you need against the Windows version before you buy. Its Pro plan is $84.99 per year on an annual plan ²⁶.

Handy is a free, open-source speech-to-text application that is listed as working on Windows, macOS and Linux ²⁷. Because it is free, it is a low-cost way to test local dictation. Expect more setup than with the commercial apps, and read the install steps and the model list before you commit to a workflow.

Local models need memory and disk space, and the larger ones are slower. Check your machine's memory against the model sizes an app offers, and test a small model first. A slow model that you keep turned on is worse than a fast model that handles your common words well.

The price chart below puts the paid options side by side. The numbers are one year of list price for one user. They are not a total cost of ownership, because the Dragon license is a one-time price and the others are recurring.

The chart shows that the Dragon options cost much more than the everyday apps. Whether that is worth it depends on the command control and vocabulary you need, not on the headline price.

Test accuracy on your own vocabulary, latency, app compatibility, privacy and total cost before you pay

Accuracy depends more on the microphone and the room than on the brand. Word error rate (WER) is a standard measure of the discrepancy between a machine transcript and a human one ²⁸. In a 2017 study, close-talking Wall Street Journal sentences recorded in a studio had a WER of 3.7% with a neural model trained on clean data ²⁹. With a single distant ceiling microphone in an apartment, using real recordings, the same kind of recognizer had a WER of 12.0% ³⁰. The same app can behave very differently at arm's length and at your desk. Test with the microphone you will use every day, not with a headset you use only for demos.

Independent evaluation of commercial speech services is limited. A 2024 study measured eleven common automatic speech recognition services on university lecture recordings ³¹ and concluded that they lack reliability in accuracy ³². Our search found no controlled 2025 or 2026 test that compares Windows voice typing with Dragon on the same audio. The comparisons online are mostly vendor blogs, which rank their own products. Treat them as leads to check, not as results.

Speed and accuracy are separate measures. A 2022 mobile text-entry study found that speech recognition was faster than typing, but the error rate for typing was lower ³³. A fast draft with more errors can take longer to finish than a slower one with fewer, so time the whole job, corrections included.

Run this short test before you pay for anything:

1. Dictate 500 words of your real work, with the microphone and room you will use.
2. Include your 20 hardest terms: names, product codes, medical or legal words, and abbreviations. Note which ones come out wrong.
3. Count the words you had to correct, and time the whole job, not just the speaking.
4. Check the app in each program you use, including your email client and browser.
5. Read the privacy setting, any storage option and any training option, and note the default.
6. Add up the cost for 12 months, including any free-tier limit you will hit.

Most apps have a free tier or trial, so you can run this test before you buy. If an app cannot handle your terms, the price does not matter. Look for a custom vocabulary feature. The custom vocabulary entry explains what to look for.

Medicine, law and other regulated work need vocabulary and compliance, not just accuracy

Regulated work adds two requirements: a large specialist vocabulary and a clear answer about where data goes. Dictation for medical notes has a long history of error-rate comparisons. In a 2000 peer-reviewed study of physicians dictating medical reports, Dragon NaturallySpeaking Medical Suite 3.0 had a vocabulary recognition error rate of 14.1 to 15.2 percent ³⁴. Those versions are long out of date. The lesson still applies: results depend on the product, the vocabulary and the speaker, so run your own test.

For health information in the United States, HIPAA's Privacy Rule allows a covered entity to let a business associate create, receive, maintain or transmit protected health information only when the entity obtains satisfactory assurance that the associate will safeguard it ³⁵. Whether a dictation app meets that standard for your practice is a legal question for your compliance team, not something this guide can settle. Ask the vendor for its written terms, and check where your audio is stored before any patient or client information goes into the app.

For a fuller treatment of professional tools, see professional dictation software. For the words these tools depend on, see the custom vocabulary and cloud dictation entries.

Frequently asked questions

The questions people ask most often each have a short answer on their own page:

• What is the best free dictation software for Windows?
• Is Windows dictation any good?
• Is Windows voice typing as good as Dragon?
• Does Dragon NaturallySpeaking work with Windows 11?
• Which AI dictation apps work on Windows?

If you are choosing now, start with the free tools for a week, run the test above on your real writing, and pay only for the gap you can measure. For the Windows tools themselves, see Windows voice typing, Windows dictation privacy and the overview of dictation on Windows. You can also try free dictation in the browser to see how a recognizer handles your voice before you install anything.