Updated August 30, 2026 · By Sumbat.T

Windows Speech Recognition in 2026: What Replaced It

What actually happened to Windows Speech Recognition

For close to two decades, Windows Speech Recognition was the answer to the question of how you use a Windows PC without a keyboard. It shipped with Windows Vista in 2007, survived every release since, and built up a genuinely devoted user base: people with repetitive strain injuries, people with motor disabilities, writers who preferred talking to typing, and a long tail of users who had spent years training it to understand them specifically.

Then Microsoft retired it. The wording in Microsoft's own support documentation is precise and worth quoting exactly, because almost every secondhand account of this change gets some part of it wrong:

Two things in that sentence explain most of the confusion you will find in forum threads. The first is the version boundary: this only applies to Windows 11 22H2 and later. If you are running Windows 10, WSR is still on your machine, still works, and will keep working. That is why you will find one person insisting the feature is dead and another insisting they used it this morning, both of them correct.

The second is the word "replaced". Voice access is not a renamed WSR. It is a different application, built on different speech technology, with a different command vocabulary and a different setup process. Things you knew how to do stopped working in ways that felt arbitrary, and the most valuable thing many long-term users owned, their trained speech profile, simply did not come with them.

The Win+H voice typing bar on Windows 11. This is the feature most people find first, and it is not the WSR replacement. Source: Windows 11 desktop.

Three built-in tools, constantly confused for each other

The single biggest source of frustration here is that Windows now has three separate speech features with overlapping names, and search results treat them as synonyms. They are not. Knowing which one you are being told to use saves an enormous amount of time.

The practical rule: if you want to write with your voice, press Win + H and start talking. If you cannot use a keyboard or a mouse at all and need to drive the whole machine, set up voice access. The two are solving different problems and the advice that treats them as one feature is the reason so many people conclude Windows dictation is broken when it is simply the wrong tool for what they wanted.

If the built-ins keep falling short

Dictation that arrives already formatted

BlabbyAI types into any Windows text field on Ctrl+Space, with punctuation, casing and paragraph breaks handled for you. 60 credits a week to try it, no card and no account setup.

Add BlabbyAI to Chrome

Every Windows speech option, side by side

Here is the whole field in one table, including where the built-ins stop and paid tools begin.

Tool	Status	Shortcut	Offline	Price	Best for
Windows Speech Recognition (WSR)	Replaced on Windows 11 22H2+	Windows + Ctrl + S (Windows 10)	Yes	Free, built in	Windows 10 machines, where it still ships
Voice access	Current, the WSR successor	Windows + Ctrl + S (Windows 11)	Yes, after model download	Free, built in	Hands-free control of the whole PC
Voice typing (Win+H)	Current, separate feature	Windows + H	No, cloud based	Free, built in	Quick dictation with zero setup
BlabbyAI	Current	Ctrl + Space	No, cloud based	Free tier, $8.49/mo unlimited	Formatted, corrected text in any app

Voice access: setting up the replacement

If you were a WSR user and you want the closest thing to what you had, this is it. Setup takes a few minutes and has one step that surprises people.

1. Press Win + U to open Settings, then go to Accessibility and then Speech. You can also just type "voice access" into the Start menu.
2. Turn the voice access toggle on.
3. Windows downloads a speech model for your language. This step needs an internet connection, even though the recognition itself runs locally afterwards. This is the step that catches people out, because the feature is advertised as working offline and the first run does not.
4. Pick your microphone and finish the short setup.
5. Say "voice access wake up" to start listening, or click the microphone on the floating bar.

Which commands survived, and which did not

If you had the old command list in muscle memory, most of it still works. The table below maps the jobs people actually did in WSR onto what voice access expects. The differences are smaller than the upheaval suggests, with two significant exceptions at the bottom.

The last two rows are the ones that matter. Losing the speech profile is the change long-time users feel most sharply, and it is worth being clear about why it is not the loss it appears to be: voice access does not need one. The old profile-based approach existed because speech models of that era genuinely could not handle an arbitrary voice without adaptation. Modern models are trained on vastly more varied speech, so they arrive already able to understand most people. You lost the training, but you also lost the need for it.

The final row is the one nobody talks about, and it is the real ceiling on both tools. In WSR and in voice access alike, the command vocabulary is fixed. Microsoft decides what the phrases are. If the thing you naturally say is not on the list, it does not get obeyed, it gets typed into your document as text. Anyone who has watched "new paragraph" appear as the literal words "new paragraph" in the middle of a sentence knows exactly how this feels.

Where the built-in tools stop

Both built-in tools transcribe. Neither of them writes. That distinction sounds like hair-splitting until you have dictated a few hundred words and gone back to read them: the text is broadly what you said, in one long undifferentiated run, with filler words intact, the false starts preserved, and the paragraph breaks you assumed would be inferred simply absent. Getting from that to something you would send takes a pass of manual cleanup that eats most of the time dictation was supposed to save.

This is the gap dedicated dictation tools exist to fill, and it is where BlabbyAI sits. It runs on Whisper large v3 turbo, which handles punctuation, casing and sentence structure as part of transcription rather than as an afterthought. You press Ctrl + Space anywhere in Windows, speak, press it again, and the text lands in whatever field your cursor was in. Word, Outlook, Teams, Chrome, an IDE, a hospital EHR, anything with a text field.

The BlabbyAI toolbar sits above the Windows taskbar and tucks away when not in use. Recording starts on Ctrl+Space from anywhere in Windows.

The fixed-vocabulary problem, solved differently

Remember the last row of the command table: neither WSR nor voice access lets you add a command of your own. BlabbyAI takes the opposite approach, and it is the most useful thing to understand about it if you are coming from years of built-in Windows dictation.

A mode is an instruction you write in plain English that runs over your speech after it is transcribed. Because the instruction is free-form and the AI receives your full literal transcript, spoken commands become something you define rather than something you look up. You write the rule once and it applies to every recording after that.

A mode is a plain-English instruction plus a keyboard shortcut. This is where you write the rules the built-in command lists never let you add.

The offline question, answered honestly

This is the one area where the built-in tools have a real and unmatched advantage, and it deserves saying plainly. Voice access runs on-device speech recognition and works with no internet connection at all, once the language model has been downloaded. WSR did the same. If you dictate on a plane, on a train through tunnels, in a facility with no network, or on a machine that is deliberately kept off the internet, the built-in tool is the one that works.

Cloud-based tools, and that includes Win+H voice typing as well as BlabbyAI and every other AI dictation app, need a connection for every recording. The reason is unavoidable: the models that produce properly punctuated, formatted, corrected text are far too large to run comfortably on a laptop. You are trading offline capability for output quality, and which side of that trade you want depends entirely on where you work.

When none of it works: the usual causes

A large share of "Windows speech recognition is broken" reports come down to a handful of repeat offenders. Work through these before concluding the feature is at fault.

• Nothing has focus. Win + H only opens when Windows has a text field to type into. Click into the document first, then press it. This is far and away the most common cause of "the shortcut does nothing".
• The wrong microphone is default. Settings, then System, then Sound. A headset that is connected but not selected behaves exactly like a broken feature.
• Microphone access is blocked. Settings, then Privacy and security, then Microphone. Check both the global toggle and the per-app list.
• Your build predates the feature. Voice access needs Windows 11 22H2 or later. No toggle under Accessibility and Speech means Windows Update, not troubleshooting.
• A remote session is eating the shortcut. Remote desktop and virtual machine clients intercept Windows key combinations. Dictate on the host, or use a tool with a shortcut that is not Win-prefixed.
• The app draws its own text fields. Some older desktop software does not report focus to Windows the way the shortcut expects, so nothing appears even with everything configured correctly.

If you are stuck on the last two, a tool that runs its own shortcut and types at the keyboard level sidesteps both, because it does not depend on Windows reporting focus in any particular way. Our voice typing not working guide goes through the full diagnostic sequence.

If you cannot install anything on your machine

A specific and very common situation: a managed work laptop where IT blocks executable installs, a Chromebook, or a machine you do not administer. The built-in Windows tools are the obvious answer because they are already there, but if you have hit their ceiling and cannot install a desktop app, there is a third route.

The BlabbyAI Chrome extension installs from the Chrome Web Store with no executable and no admin rights. It puts a small bubble next to whatever text field you click into, and it works anywhere the browser goes: Gmail, Google Docs, Notion, Salesforce, Jira, a web-based EHR, any internal tool your company runs in a tab. It does not have History, which is a Windows-app feature, but the dictation and the custom modes are the same.

It is also the answer for anyone reading this on a Mac or a Chromebook, where the Windows tools this article is about do not exist at all.

One thing the desktop app does that nothing built in does

Worth knowing if you dictate anything you cannot afford to lose. The BlabbyAI Windows app saves your audio to disk the instant a recording stops, before transcription even begins. If the transcription fails, the network drops mid-send, or you realise afterwards that you wanted a different mode applied, the recording is still on your machine and you can replay it or run it through again.

Neither WSR nor voice access nor Win+H keeps anything. If a transcription goes wrong with the built-in tools, the words are simply gone and you say them again. History holds 500 entries by default, lives entirely on your own machine, and never touches a server.

Frequently asked questions

The bottom line

Windows Speech Recognition is gone from current Windows 11 and voice access took its place in September 2024. If you are on Windows 10 it is still there. If you are on Windows 11 and you want the successor, it is Win + Ctrl + S after enabling it in Accessibility settings, and it is better than what it replaced at the thing it does: hands-free control of the machine, offline, free.

If what you actually wanted was to write faster, press Win + H instead and skip the setup entirely. And if you have been round that loop already and the problem is not capturing words but cleaning them up afterwards, BlabbyAI is $8.49 a month with a free tier that refills weekly, and it is the only option here where you write the command vocabulary yourself. On a locked-down laptop or a Chromebook, the Chrome extension installs where an executable will not.

For more on the built-in shortcut specifically, our Win+H voice typing guide covers it in depth, and best dictation software for Windows compares the paid options with real prices. If a specific shortcut has stopped responding, voice typing shortcut keys lists what each one does.