getting started
Install the desktop app, meet YAP, learn what lenses are, and make your first recording.
yapyap is a desktop app for macOS, Windows and Linux. It records, transcribes and understands your voice notes entirely on your own machine — the only things that ever use the network are downloading models, checking your license, and (optionally) syncing recordings from your phone over your own wifi.
install
Grab the installer from yap-yap.app. Bought a license already? The customer portal always lets you re-download the latest installer for your platform.
On Linux the download button offers three packages, because one download can't suit every distribution:
| package | for |
|---|---|
.AppImage | any distribution. Nothing to install: mark it executable and run it |
.deb | Debian, Ubuntu, Mint and other apt distributions |
.rpm | Fedora, RHEL, openSUSE and other rpm distributions |
The .deb and .rpm pull in their dependencies (see
system requirements) and put yapyap
in your application menu; the AppImage expects webkit2gtk 4.1 to be there
already.
After that, the app keeps itself current: when a new version ships you get a quiet notification, and the update only installs when you say so — it will never restart the app mid-recording.
first launch
The first run is a short conversation with YAP (the mascot) — five turns, and exactly one question:
- hello — what yapyap is, and that everything stays on this machine.
- lenses — the idea the whole app is built on (below).
- brain — yapyap sizes up your machine and recommends one of three model presets. One button takes it; the download runs in the background while you carry on.
- anonymous usage counts? — the one question. Off unless you say yes.
- done — 7 days to try, no key and no email to enter.
Setup asks for no permissions. The microphone is requested when you first press record, Accessibility when you first dictate — each one in front of the thing it protects.
Details in first run. Every choice can be changed later under settings.
lenses, the short version
This is the part worth reading twice. yapyap doesn't produce a result per recording — it reads each recording through lenses, and a lens is just a reusable question you've decided is worth asking every time:
| the lens | the question it keeps asking |
|---|---|
| transcript | what was said, and by whom? |
| summary | what would someone who missed this need to know? |
| action items | who owes whom what, by when? |
Those three ship with the app. They're the starting set, not the ceiling.
What makes a lens a lens: a prompt for the local language model, a declared shape for the answer, and a view that renders it. Because the shape is declared, an answer comes back as a checklist or a table of owners and due dates — not a wall of text you have to re-read.
Three things follow, and they're the whole pitch:
- Swappable. The same recording, read different ways. Run a lens now, or flip it always-on and every future recording arrives with that insight already extracted.
- Editable. Every shipped lens is yours to change — the prompt, the shape, the view. Reset to default undoes it.
- Yours to grow. Build one from scratch, or install one from the marketplace. "Decisions made", "objections raised", "questions I didn't answer" — if you ask it of every conversation, it's a lens.
And it all runs on your machine, on your model. A lens is a question you own, answered by a model you own, about audio that never left home.
Go deeper: lenses · the marketplace · publishing your own
your first recording
Press . (period) anywhere in the app — or use the tray icon — and talk. Words stream in live. Press . again to stop. The recording then flows through the local pipeline (transcribe → who-spoke-when → meaning → insights) and lands in your library, searchable with ⌘ K.
Speak whatever language you like: transcription detects the spoken language by ear, always.