Text while you talk
Words are typed into your document as they are recognised: not into a preview bubble, not when you let go. You catch a misheard word mid-sentence.
Hold ⌥Space and talk. Words appear in any app as you speak, cleaned up on your Mac when you let go.
I think we should move the launch to Thursday. The build isn't ready yet.
No window to open, no field to click first. WunderWhisper types into whatever app has the cursor: Mail, Slack, Notes, Cursor, a browser tab.
Rebind it if you like. For long dictation, set a Start/Stop key: press once to start, again to stop. Works with remotes and foot pedals.
Words appear in the app as you say them, so you see what was heard before you finish the thought.
so the plan for Tuesday is
On-device AI drops the ums and repeats, fixes punctuation, and swaps the clean version in place.
Speech recognition runs on your Mac with Apple’s on-device model. The audio never leaves it.
Words are typed into your document as they are recognised: not into a preview bubble, not when you let go. You catch a misheard word mid-sentence.
Types where the cursor is, through the macOS Accessibility API. Mail, Slack, Notes, Chrome, Cursor, VS Code.
A small AI model on your Mac removes filler words and stutters and fixes punctuation. No API key, no internet after the one-time download.
If the cleanup adds or drops a negation, it is thrown away and your own words stay. Losing polish is better than losing meaning.
The clean version replaces your dictation only if that exact text is still right before the cursor. Otherwise nothing is touched, and Copy Last Dictation has it.
Hold ⌥Space for quick notes, or press a Start/Stop key for long dictation, up to 30 minutes. Remotes and foot pedals work too.
A small pill at the bottom of the screen shows the waveform. ✕ removes what was typed, ■ ends the dictation.
If your microphone sends no sound at all (muted, wrong input), the dictation stops after 3 seconds and names the mic.
All of dictation is free: no word limit, no account, no trial clock. It costs us nothing to run, so it costs you nothing.
Press Record in the menu bar before a call. WunderWhisper writes the transcript while you talk and the notes when you hang up. Nobody sees a recorder join.
Let’s start with the pricing page. 00:12
I’d keep the free tier. We shouldn’t cap words. 00:18
Agreed. Can we ship it by Thursday? 00:26
Thursday works if Lena reviews the copy. 00:31
Pricing page ships with an uncapped free tier.
No word cap on the free tier.
- [ ] Lena: review the copy
Does Thursday hold if the review slips?
Speech-to-text always runs on your Mac. This choice is only about the model that cleans up dictation and writes meeting notes, and it only ever sees text. Switch any time.
The model runs inside WunderWhisper on your Mac’s GPU. Nothing else to install.
Default · one model for dictation and meetings
Use your own OpenAI account. Direct request, no middleman.
Hundreds of models through one API key.
Bring any provider or server that speaks the OpenAI API.
The whole design of the app is a promise you can audit. Read the privacy policy → every network connection the app makes is listed there.
No subscription. Pro is a one-time license key, and the app lets you record 3 meetings with every feature before you decide.
macOS 26 or later · Apple Silicon · Signed and notarized by Apple · Updates itself
Yes. All of dictation is free with no word cap, no account and no time limit: hold-to-talk, Start/Stop, live typing and AI cleanup. Pro only adds meeting recording.
No. Speech recognition runs on your Mac with Apple's on-device speech model. If you choose a cloud AI provider for cleanup or notes, only the text is sent, directly to that provider with your own key.
WunderWhisper works the same way in every app, starts with a hold-to-talk hotkey, cleans up filler words and punctuation when you let go, and checks the cleanup never adds or drops a "not". Pro adds meeting recording with speaker labels and notes.
No. WunderWhisper records your microphone and the call audio on your Mac, so it works with Zoom, Google Meet, Teams, Slack huddles or any app that plays sound, and nobody sees a recorder join. Tell the people on the call you are recording.
$29, once. One license key works on 3 Macs and includes all future updates. The app lets you record 3 meetings with every feature first, and meetings you recorded stay readable whether you buy or not.
macOS 26 or later on Apple Silicon (M1 or newer). The on-device AI models run on the Mac's GPU.
The languages Apple's on-device speech recognizer supports on macOS 26. You pick one language per session in Settings; Turkish isn't supported by Apple's recognizer yet.
In a folder on your Mac (~/Library/Application Support/WunderWhisper/Meetings): the audio, the transcript and the notes as Markdown. Nothing is uploaded unless you pick a cloud AI provider, and then only text.
Typing into other apps goes through the macOS Accessibility API, which the App Store sandbox blocks. You download it from this site instead; it is signed and notarized by Apple and updates itself.
Have a question, feedback, or feature request? We’d love to hear from you.