How can we help?
Start with the basics, or reach out directly.
Installing and setting up
The download, the two permissions macOS asks for, and what to do when a build won't open.
Download & setup →How do I actually do X?
Step-by-step guides for the things people ask for most — dictating email, prompts, meeting notes.
How-to guides →Every question, answered
The full FAQ: the voice keyboard, the Notetaker, languages, accuracy, privacy and plans.
Read the FAQ →What does this word mean?
Plain definitions for the terms this site and the app use, from all-party consent to word error rate.
Open the glossary →The basics
What is WisprBee?
WisprBee is an AI voice keyboard and notetaker for macOS. You hold one key, say the thing, and let go; finished writing appears where your cursor was, already suited to whichever app is in front of you — brief in Slack, properly structured in Gmail, an engineered prompt in Cursor, a filed ticket in Linear. The one key works in every text field on the machine, across 75 spoken languages, and the text arrives under 2.5 seconds after you let go. In a meeting, the same key runs the Notetaker.
How is WisprBee different from other dictation tools?
Most tools hand you a raw transcript and leave the tidying to you — or make you pick a format up front. WisprBee reads whichever app is in front of you and writes to suit it, unprompted. There is no mode to choose.
What does “voice keyboard” mean?
A dictation tool converts speech to text. A voice keyboard does the whole job the keyboard does: it produces the finished text you would have typed, formatted for where it's going. That is the difference between getting words and getting writing — a transcript still needs punctuation, filler removal and reformatting before you can send it; a voice keyboard's output is already sendable.
Does WisprBee replace my keyboard?
No — it sits alongside it. WisprBee runs at the OS level and only does anything while you hold your chosen push-to-talk key. Your keyboard, trackpad and every other shortcut keep working exactly as before, and you can type over anything WisprBee pastes.
Which AI models does WisprBee use?
Speech-to-text and text polishing are routed through OpenRouter to models from OpenAI and Google: an OpenAI speech model transcribes English and other Latin-script languages, and Google models handle other languages and the text polishing. WisprBee picks whichever engine is most accurate for the language you're speaking. If those services are unreachable, it can fall back to an on-device model so dictation still works.
Who is WisprBee for?
Anyone who writes more than they'd like to type. It's built for people who move between many apps in a day — founders and managers clearing Slack and email, writers drafting out loud, support teams answering tickets, students taking notes, and multilingual professionals who think in one language and need to write in another.
All the questions — the voice keyboard, the Notetaker, languages, privacy and plans.