1. Two cloud recognizers: what Microsoft and Wispr each say
Microsoft’s voice typing page says voice typing “uses online speech recognition, which is powered by Azure Speech services,” and that you need an internet connection, a working microphone, and the cursor in a text box. Press Windows key + H, wait for the “Listening...” alert, and say “Stop listening” or select the microphone to finish. On Windows 10, the same shortcut opens what Microsoft calls dictation, and that section also says the PC must be connected to the internet.
Wispr’s Data Controls page says Flow’s transcription always occurs in the cloud, and its help center says an internet connection is required. So if privacy from the cloud or dictation on a plane is the reason you are comparing, neither option solves it. Microsoft’s offline tool is voice access, a Windows 11 feature for controlling the PC and writing text by voice that, once its language files have downloaded during setup, works without an internet connection.
2. Price, hardware, and setup on a Windows PC
Voice typing is part of Windows 10 and 11, so there is nothing to buy, install, or sign up for. Microsoft lists only a microphone, a connection, and a focused text box as requirements. The optional Fluid dictation, which corrects grammar, punctuation, and filler words with small language models on the device, is limited to Copilot+ PCs in English locales; its models download in the background, and the setting stays dimmed until they finish.
Wispr Flow is a download that needs Windows 10 or 11 on an x64 processor, with 8 GB of RAM recommended and about 200 MB free. Windows on ARM and Snapdragon PCs are refused with an “Unsupported Windows Architecture” message. Setup asks only for microphone access, through Windows Settings, and every user needs a Wispr account. As of October 3, 2026, Free costs $0 and caps desktop dictation at 2,000 words a week, while Pro costs $15 a month or $12 a month billed annually, and students and educators get 50% off Pro.
- Voice typing: $0, built into Windows 10 and 11.
- Wispr Flow Free: $0, 2,000 desktop words a week, x64 PCs only.
- Wispr Flow Pro: $180 a year paid monthly, or $144 billed annually.
3. Thirty-six languages against a hundred, and how text gets tidied
Microsoft lists 46 language and region variants for voice typing on Windows 11, covering 36 languages from Bulgarian to Vietnamese, with six kinds of English. You change the voice typing language by changing the Windows input language, with Windows key + Spacebar or the taskbar language switcher. Out of the box, tidying is modest: automatic punctuation, a profanity filter, and spoken commands such as “delete that”, “select that”, and “undo that”, with Fluid dictation adding grammar and filler-word fixes on Copilot+ PCs.
Wispr advertises more than 100 dictation languages. Flow cleans up the transcript before inserting it and adjusts tone to the app you are in; Wispr’s own examples are a formal tone in email and a casual one in messages. It adds a dictionary for names and jargon, snippets, a hands-free mode, and Command Mode for editing by voice. With Auto-add to Dictionary on, Flow watches the box it pasted into and learns spellings you correct, and the default hotkey on Windows is Ctrl+Win.
4. Microsoft’s voice clips and Wispr’s data switches compared
Microsoft’s speech privacy page says voice data for online speech recognition is sent to Microsoft only to provide the service and create transcriptions, and that Microsoft does not store, sample, or listen to voice recordings without your permission. Contributing voice clips to improve recognition is optional and managed from voice typing’s settings; with a work or school account, sharing voice data is not available at all. Online speech recognition can be switched off under Settings → Privacy & security → Speech, which leaves only device-based features, such as Narrator, available.
Wispr’s controls are separate switches. “Improve the model for everyone” lets Wispr use your audio, transcripts, and edits to train models; its help center says the switch is on by default for Free and Pro and can be turned off, while Growth and Enterprise keep it off. Dictation Cloud Storage decides whether transcripts, audio, and history stay on Wispr’s servers. Wispr also uses the app name and nearby text-box content for formatting, may collect usage statistics such as word counts whatever your settings, and encrypts data in transit and at rest.
- Voice typing: no recordings kept unless you opt in to voice clips.
- Wispr Flow: turn off model training yourself on Free and Pro if you do not want it.
- Both: recognition happens on someone else’s servers.
5. PowerShell, Windows Terminal, and WSL tabs
Voice typing’s English commands include “press Enter”, “press Tab”, and “press Backspace”. In a document, “press Enter” starts a new line; at a PowerShell or cmd prompt it runs whatever is on the line, and Tab can trigger completion that changes it. Automatic punctuation can also leave a period at the end of a command. For a sentence to a coding agent that is harmless; for shell syntax, turn automatic punctuation off in the flyout settings or type the command yourself.
Wispr lists Command Prompt, PowerShell, Windows Terminal, and WezTerm as places where Flow pastes directly. In WSL windows, Linux virtual machines, and remote SSH sessions it still transcribes but cannot paste, so you press Paste last transcript (Alt+Shift+Z) or Copy last transcript (Alt+Shift+X). Wispr warns that context-aware formatting can transform shell commands, and that if the target app runs as administrator, Flow must run as administrator too.
- Win+H at a shell prompt: never say “press Enter” before reading the line.
- Wispr Flow in a WSL tab: dictate, then press Alt+Shift+Z.
- Elevated terminal: Wispr says to run Flow as administrator as well.
6. A third option if you need offline Windows dictation: IraVoice
Labelled plainly, since IraVoice publishes this page: if the reason you are comparing these two is that both need the internet, IraVoice is a local alternative. IraVoice 1.2.2 runs on Windows 10 or 11, 64-bit (x64) and recognizes speech on the PC with NVIDIA Parakeet; the default Parakeet Unified EN model, about 700 MB, downloads once, and dictation then works with no connection. Hold Right Ctrl, speak, and release; Esc cancels, and IraVoice does not press Enter, so you can read a command before it runs.
It pastes into the focused window, so text lands in Windows Terminal tabs running WSL or SSH. In Windows Terminal, PowerShell, cmd, VS Code, Cursor, and JetBrains IDEs it spells names from your Git project, though not reliably inside WSL. When the focused app runs as administrator, Windows blocks its input, so IraVoice leaves the text on the clipboard and says so. Its limits: the default model is English (other offline models cover 25+ languages), there is no phone app, it still connects for the model download, license check-ins, and update checks, and IraVoice Local costs $79 once for one computer, $109 for two, or $149 for three after a 3-day free trial.