Speech to text on Windows 11

Path 1: press Win+H and start talking

Prerequisites: Windows 11, an internet connection, a working microphone, and your cursor in a text box. No download or account. Time: under a minute.

Expected result: the toolbar reads Listening, and the words you say show up at the cursor as you speak. Turning automatic punctuation on (the gear icon on the toolbar) lets it add commas and full stops based on what you say.

If it does nothing: voice typing uses online speech recognition powered by Azure, so it needs an internet connection, a working microphone, and a cursor sitting inside a text box. No connection, no transcription. The full fix list is two sections down.

Coming from an older tutorial that talks about Windows Speech Recognition? That feature was replaced by Voice Access for Windows 11 22H2 and later back in September 2024. The old WSR control panel only lives on in older Windows versions now. So if a guide tells you to open a Speech Recognition wizard and you cannot find it, the guide is out of date, not your PC.

When the built-in voice typing is enough

I am not going to tell you to install software you do not need. For a lot of tasks, Win+H is the right answer, and it costs nothing.

Use the built-in tool when the dictation is short, you have a connection, and the stakes are low. A reply to a coworker on Teams. A quick note in OneNote. A search query you would rather speak than thumb out. It handles automatic punctuation, and it works in any standard text box across Windows 11. For 30-word bursts where you are online anyway, opening a second app would be slower than just talking.

A second built-in feature confuses people, so let me clear it up. Voice Access is not the same thing as voice typing. Voice Access lets you control the whole PC and author text by voice, and unlike Win+H, it runs offline using on-device speech recognition after a one-time language download. It needs Windows 11 version 22H2 or later. So if you need full hands-free PC control (clicking, scrolling, opening apps by voice), Voice Access is the built-in tool to reach for, not voice typing. Different jobs.

Where Win+H falls short (offline, accuracy, languages)

The built-in tool has three real ceilings. None of them are dealbreakers for a quick note. All three start hurting the moment you do longer or more serious work.

The privacy angle is the one I think about most. Your dictation — the email to your kid's school, the draft of a contract, the half-formed idea you would never say out loud in a meeting — leaves your machine and goes to a server. For a Teams running five late that is nothing. For the things you care about, it is worth knowing where the audio goes.

Win+H not working? The three usual culprits

When Win+H does nothing, it is almost always one of three things. Check them in this order, because that is how often each one is the cause.

If none of those land, the deeper issue is usually a language pack that did not finish installing, or a Windows update mid-flight. That is also the point where I stop fighting the built-in tool and set up something I control end to end, which is Path 2 below. If it keeps failing after that, we wrote a separate guide to voice typing not working on Windows with the longer checklist.

Path 2: set up a dedicated dictation app

Whisper is the desktop app I build, and it does the three things Win+H cannot: it transcribes offline on your own CPU, it works through one system-wide hotkey in any application, and it lets you pick the engine for your hardware and languages instead of one fixed cloud model. Here is the full setup, start to finish.

Prerequisites: Windows 11, about 1 GB of free disk for a mid-size model, a microphone, and a free account (no payment method to start). You need a connection for the one-time download only; transcription afterward is offline. Time: 5 to 10 minutes, most of it the model download.

Expected result: with the model downloaded, you hold Ctrl+Space in any application, speak, release, and your words paste at the cursor with no internet in the loop after the download. Saying Hey whisper triggers an AI clean-up pass on the text before it lands, if you turn that on.

If the hotkey misfires: rebind it in Settings. I learned this one the hard way. The first version of the hotkey handler fired the recording-stop callback six times per real keypress on Windows, because the Windows input framework generates phantom Ctrl+Space release events at unpredictable intervals. It worked on a clean machine and broke on any laptop with a second language input enabled. It took telemetry, a 50ms guard that was not enough, and finally a 300ms debounce that was. My daughter's verdict stands: this is why dad's emails take forever.

On language coverage, the local Whisper engine handles 99 languages on its multilingual models, while the .en builds are English-only and a touch faster for that one job. The Parakeet engine runs 5 to 10 times faster than Whisper on CPU and covers English plus 24 European languages (25 in total), though it skips Asian languages and translate-to-English. If offline-first is the part that matters to you, our deeper guide to offline speech to text covers the engines in more detail.

Win+H vs Voice Access vs a dedicated app

Three ways to put speech into text on Windows 11, side by side. The table covers only what each tool documents. No invented accuracy or speed scores.

If your only need is a quick Teams reply while you are online, Path 1 wins on simplicity. It is already on your PC. The moment offline, any-app coverage, or a missing language enters the picture, Path 2 earns the install.

Local vs cloud: which Whisper mode for your PC

Whisper runs in two modes, and the choice comes down to your hardware and whether you want web access.

Local mode does everything on your machine. Pick a Whisper model sized to your PC: Base is around 140 MB and runs on almost anything, Small is around 480 MB, Medium around 1.5 GB, and the multilingual Large v3 is around 3 GB for the best accuracy if you have the RAM. Or pick Parakeet at around 600 MB for the fastest local option if you mostly work in English or European languages. None of it touches the internet after the download.

Cloud mode is the escape hatch. It uses your own OpenAI key: transcription through gpt-4o-mini-transcribe or gpt-4o-transcribe, and web search when you want a live answer pasted at the cursor. You bring the key; we take no cut.

Here is the opinion I will stand behind: try local mode first. If your Windows PC is from the last four years, you do not need cloud for everyday dictation, and local mode keeps your audio on your machine where it belongs. Cloud is the fallback for when you hit a wall, not the default. Whisper is free for the entire local pipeline once you sign in, with no payment method required to start. The Cloud surface is the paid Pro tier. Details are on the pricing page. For the full local walkthrough, the voice to text on Windows guide goes step by step.

When to skip the dedicated app

I would rather you keep Win+H than install something you will not use. Skip a dedicated app, and stay on the built-in voice typing, when all of these are true:

• You dictate short bursts, not long documents.
• You are online whenever you dictate.
• You work only in a language Microsoft's voice typing already covers.
• Your audio is low-stakes and you do not care whether it leaves the machine.

Win+H is free, already installed, and good at exactly that job. The dedicated app earns its place the moment you cross one of those lines — a plane, a contract draft, a language Microsoft skipped, or any app that is not a standard text box.

Honest pricing

Whisper's local mode is free for everyone who signs in: Whisper and Parakeet transcription, AI enhancement, history, presets, custom hotwords, hardware acceleration, model downloads, and the global hotkey, all of it, with no card required to start. Whisper Pro adds the Cloud features on top: OpenAI cloud transcription, cloud AI enhancement, and voice web search. The built-in Windows voice typing is also free, because it is part of Windows. The full plan breakdown lives on the pricing page. I would rather you read the exact numbers there than trust a figure in a blog post that drifts out of date.

Two paths, one decision. If you are online, the note is short, and the stakes are low, press Windows key + H and talk — it is free and already on your PC. The moment you need it on a plane, in any app, in a language Microsoft skipped, or with your audio staying on your own machine, set up the dedicated app instead. I crossed that line somewhere around the third meeting note I dictated one-handed while making lunchboxes, and I have not typed a long email since.