How to Set Up Wispr Flow Voice Dictation
Wispr Flow turns speech into formatted text in any app on your computer, and it works the way dictation should have always worked: hold a hotkey, talk normally, release, and clean text appears where your cursor is. The free tier gives 2,000 words a week.
What is Wispr Flow?
Wispr Flow is a voice dictation app from Wispr AI, founded in 2021 by Tanay Kothari and Sahaj Garg. It runs as a background app on macOS and Windows, listens while you hold a key, and inserts text into whatever field is focused — Gmail, Slack, Notion, VS Code, a terminal.
The difference from system dictation is formatting. Say “so I think we should push the launch to Tuesday, actually no, Thursday, and tell the team” and you get “I think we should push the launch to Thursday. Let’s tell the team.” It removes filler words, fixes self-corrections, and adds punctuation.
That self-correction handling is the actual product. Every other feature is polish around it.
How do you set up Wispr Flow?
Installation is quick, but the permissions are where people stall.
- Download from wisprflow.ai and install. macOS 13+ or Windows 10+.
- Grant Microphone and Accessibility permissions. Accessibility is mandatory — without it, the app cannot type into other windows, and it will appear to record but insert nothing.
- Create an account. Free tier is 2,000 words per week; Pro is about $12/month billed yearly.
- Set your hotkey. Default is Fn on Mac; most people change it to something that does not conflict with their window manager.
- Test in a scratch document first. Click into it, hold the hotkey, say two sentences, release.
The first-run tutorial is worth doing once. It calibrates to your speech and adapts the formatting, and skipping it means a week of odd punctuation.
Windows users: the app is more recent on Windows and some per-app behavior differs. Apps that intercept keystrokes at a low level — some remote desktop clients, a few IDEs with custom key handling — reject the injection. Test your top three apps before committing.
How well does the formatting actually work?
On English, very well. Filler removal, punctuation, and self-correction are reliable enough that most users stop proofreading sentence structure and only check names and numbers.
Known weak spots:
- Proper nouns and product names get mangled. “Kubernetes” and “Figma” mostly work; internal project codenames do not.
- Numbers sometimes arrive as digits, sometimes as words, depending on context, and there is little you can do about it.
- Long dictation past about 90 seconds starts to accumulate small errors.
- Non-English languages work but with noticeably more errors; German compound nouns are a particular struggle.
The workaround is a personal dictionary. Add your company’s proper nouns once and it stops being an issue for the words you use daily.
What does it cost in 2026?
| Plan | Price | Words |
|---|---|---|
| Free | $0 | 2,000 words/week |
| Pro | About $12/month yearly (or $15 monthly) | Unlimited dictation, all apps |
| Team | Custom | Shared billing, admin controls |
2,000 words a week is about 25 minutes of talking per day — enough to know if it fits your workflow and not enough to replace typing.
The Pro price is comparable to a streaming subscription and, subjectively, saves more time than one. If you write 3,000 words a day, that math is easy.
Where does it fall short?
- Verification is still required. Anything with a number, a name, or a URL needs a glance before sending. A wrong figure in an invoice email is expensive.
- Noisy environments degrade it badly. Open-plan offices and cafes produce garbled output; a decent headset mic fixes most of it.
- Not for public or shared spaces — talking to your computer in a meeting room is its own problem.
- Context awareness is shallow. It reads a little of the surrounding text but does not know your style guide or your teammate’s name spelling.
- Blocked in some corporate environments because accessibility permissions look like keylogging to endpoint security tools. Expect an IT conversation.
- Tone control is limited to a few presets — formal, casual, very casual — not per-app personality.
- No offline mode; dictation requires a network round trip.
Who should use Wispr Flow?
Anyone who types more than a thousand words a day and thinks faster than they type: support agents, founders answering email, developers writing commit messages and docs, writers doing first drafts. It is especially good for people with RSI or wrist pain, and for anyone whose job is mostly conversational replies.
Skip it if your work happens in quiet shared spaces, if you mostly type code rather than prose (dictating syntax is genuinely bad), or if your IT policy forbids accessibility-permission apps. Also skip if you send a lot of numbers — the verification overhead eats the time saved.
Quick Summary
- Wispr Flow is a 2021-founded dictation app for macOS and Windows that inserts formatted text into any focused field.
- Free tier: 2,000 words per week; Pro is about $12/month billed yearly.
- Its edge is removing filler and handling spoken self-correction, not raw transcription speed.
- Requires Accessibility permission — corporate endpoint security often blocks it.
- Always verify names, numbers and URLs before sending.
