Speech to text on Windows.

Pick the tool by whose speech it is

If you came here to dictate, start with voice typing. It costs nothing and writes into whatever you are working in, as does voice access. Both listen to your microphone, not to what your PC plays, and Microsoft says voice typing filters out sound from your speakers on purpose. For a call or a video, you need a tool that listens to the PC itself.

Type with your voice: voice typing

1. Click where the text should go

   Put the cursor in a text box: an email, a document, a chat, a search bar. Voice typing only starts when the cursor is in a text box.

2. Press Windows key + H

   The voice typing menu opens and starts listening. Wait for Listening..., then talk. On a touch screen, tap the microphone on the touch keyboard instead.

3. Say the punctuation, or let Windows add it

   Say "period", "comma", "question mark" or "new line". Or select the gear in the voice typing menu and turn on Automatic punctuation.

4. Fix mistakes and stop by voice

   Say "delete that" to remove the last thing you said, "undo that" to undo the last change, and "stop listening" when you are done. Pressing Windows key + H again stops it too.

Word, Outlook and PowerPoint in Microsoft 365 have their own Dictate button as well. Microsoft says it needs a microphone and a reliable internet connection.

Voice typing settings, languages and privacy

• The gear in the voice typing menu holds the settings. The voice typing launcher starts it quickly from a text box. Wait time before acting sets how long it waits before carrying out a command. The others are Automatic punctuation, Filter profanity and the default microphone.
• Microsoft lists 46 voice typing languages and regional versions, among them English, Spanish, French, German, Portuguese, Russian, Hindi, Japanese, Korean and Chinese. To dictate in another language, add it under Settings > Time & language > Language & region, then switch to it with Windows key + Spacebar.
• Your voice goes to Microsoft. Its privacy page says online speech recognition sends voice data to Microsoft to create the text. It also says Microsoft does not store, sample or listen to recordings without your permission. Sharing voice clips to improve the service is optional, and the voice typing settings link to where you start or stop it.
• Microsoft says that with Online speech recognition turned off, under Settings > Privacy & security > Speech, only features that work on the device are left. Voice typing is not one of them.
• On a Copilot+ PC, fluid dictation fixes grammar, punctuation and filler words as you speak, in English. Microsoft says that part runs on small language models on the device. Say "undo that" to get your own words back.
• Voice typing hears only you. Microsoft says it suppresses audio coming from your speakers, so it will not type out a video or the other side of a call.

Speech to text with no internet: voice access

1. Turn it on

   Go to Settings > Accessibility > Speech and turn on Voice access, or search for voice access from Start. It is in Windows 11, version 22H2 and later.

2. Download your language once

   Pick a language and select Agree and continue. Voice access downloads the files it needs to recognize speech on the device. That one step needs an internet connection.

3. Pick a microphone and talk

   When the voice access bar appears at the top of the screen, click into a text box and speak. Say "voice access sleep" to pause it and "voice access wake up" to carry on, or press Alt + Shift + B.

Voice access does more than voice typing. You can open apps, click buttons and edit text by voice. It covers fewer languages. Microsoft's setup page lists English, Spanish, French, German, Italian, Japanese and Chinese. Its settings let you add names and unusual words to its vocabulary, and on a Copilot+ PC fluid dictation works here too, in English.

Windows Speech Recognition, the older tool, has been replaced. Microsoft says voice access took its place in Windows 11, version 22H2 and later, in September 2024, and that it is still in older versions of Windows, such as Windows 10.

When voice typing is not working

• Nothing happens when you press Windows key + H. Click into a text box first. Voice typing does not start anywhere else.
• It opens but types nothing. Check the internet connection, then check that Online speech recognition is on under Settings > Privacy & security > Speech.
• It does not hear you. Pick the right microphone in the voice typing settings, and check that microphone access is on under Settings > Privacy & security > Microphone.
• The words come out in the wrong language. Voice typing follows your input language. Switch it with Windows key + Spacebar.
• A message tells you to use voice access. Voice access is already running, and it types too. Dictate with it, or turn it off first.

When the speech is not yours: calls, videos, a lecture

Windows has one built-in tool for this: Live Captions. Press Windows key + Ctrl + L and it captions anything the PC plays, on the device, in 10 languages. It is good for following along, but it is not a transcript. Microsoft says generated captions are not stored anywhere, so when the call ends, the words are gone. The live captions guide linked at the foot of this page covers its settings and fixes.

Fraze is a Windows and Mac app for the rest. It listens to system audio, which means a Teams, Zoom or Meet call in any app, a YouTube video or a lecture stream. It can also listen to your microphone, or to both at once. Words appear while the speaker talks and settle when they pause. A small captions window can stay above your other windows. Fraze keeps each finished session in History, ready to copy or save as a text file.

It all runs on your PC. No audio or text leaves it, no bot joins the call, and once a language is downloaded it works with the internet off.

Write down a call or a video with Fraze

1. Install Fraze

   Download it for Windows 11. No account is needed, and the first 30 minutes are free.

2. Choose what it listens to

   Click the dark audio pill. Check System audio for a call or a video playing on your PC, Microphone for people in the room, or both, which the pill shows as Mix.

   

3. Pick the spoken language

   Click the language pill, set Spoken language, and set Translate to No translation. The first time, Fraze downloads that language. After that it needs no internet.

4. Press Start listening

   Ask the people on the call first. Then start, and the words appear as they are spoken. The captions button turns Fraze into a small window you can keep above the call.

5. Keep the text

   When you stop, Fraze saves the session in History on your computer. Copy it, save it as a .txt file, or delete it.

What to expect from Fraze, honestly

• It does not type into other apps and has no voice commands. To write an email by voice, use voice typing or voice access. To use Fraze text somewhere else, copy it from History.
• It is live only. It cannot open an audio or video file. You can play a recording and let it listen, at the speed of the recording.
• The saved text is plain lines, with no speaker names and no timestamps. Fraze keeps no audio.
• On Windows it hears 25 European languages, among them English, Spanish, French, German, Italian, Portuguese, Polish, Dutch and Ukrainian, plus Chinese, Japanese, Korean, Cantonese, Vietnamese and Thai. It does not hear Hindi or Arabic. Voice typing covers Hindi.
• It can also translate what it hears into English and many other languages, if you want that.
• The captions window stays above other windows, but not over an app in full screen.
• It needs Windows 11 on x64 with a 4-core AVX2 processor and 8 GB of memory. It does not run on Windows 10.

What it costs

Voice typing, voice access and Live Captions come free with Windows. Fraze gives you the first 30 minutes free, with no account and no card. After that it is $29.99 once or $2.99 a month, both unlimited. The work happens on your PC, so there is no per-minute charge, and a two-hour lecture costs the same as a short call.

Questions