You speak.
It types.
Anywhere.
Words become
typed. Everywhere.
Hold ⌥ Space. Speak naturally — pauses, fillers, accents, all of it. Wasper transcribes in the active app the moment you let go. Slack, Mail, Notion, VS Code — if you can type there, you can talk there.
If you can type there,
you can talk there.
Wasper installs as a system accessibility helper, so it works inside every macOS text field — no per-app extensions, no plugins, no API integrations. App-specific style presets ship for the 78 most common ones.
Slack Casual 
Mail Formal 
VS Code Technical function authenticate()
Telegram Translate 
Notes Lists send invoice to Acme
each with its own writing style
And it stays
on your Mac.
Dictation, translation, and AI rewriting all run on Apple Silicon. Bundled local models. No audio leaves your device, no transcript leaves your device, no prompt leaves your device. Unless you choose otherwise.
Microphone
Wasper
Transcribe · Translate · Rewrite
Active app
Audio stays put. Recordings stream into the local speech engine. Nothing is written to disk by default.
Transcripts stay put. Stored locally on your Mac. Searchable, exportable, never uploaded.
No content telemetry. We log app launches and crashes only — never the words you say.
Cloud is opt-in. Bring your own Claude / OpenAI / Gemini key if you want. Off by default, off in two clicks.
Until
it adds up.
Every dictation, every rewrite, every translation — quietly counted. Months from now you'll see the hours you got back, the words you spoke instead of typed, the apps that finally became thumb-free.
Talk less.
Ship more.
* Measured in-house on a MacBook Pro (M1 Pro). With chunked Parakeet on the Apple Silicon GPU, transcribed text is pasted ~1 second after you stop speaking — and stays that fast regardless of how long you talk; engine throughput is ~72× real-time. Real-world speed varies with your Mac, audio length, language, and settings.