first run
A five-turn conversation with YAP — what it asks, what it doesn't, and what it sets up.
First run isn't a wizard with a Next button. It's a short conversation with YAP, the mascot, and it contains exactly one question.
Nothing here is final — every choice has a mirror under settings.
what it never asks for
Setup asks for no system permissions at all. Every grant is asked for at the moment it's needed instead:
| grant | asked when |
|---|---|
| microphone | you start your first recording |
| Accessibility (macOS) | you first use dictation |
| system audio (macOS) | you first record what your machine plays |
| local network (macOS) | yapyap first advertises itself to your phone |
A permission dialog makes sense standing in front of the thing it protects. In a welcome thread it's a customs desk.
the five turns
1. hello
Who yapyap is, in two lines: it listens, writes down what was said, and keeps all of it on this machine. Your audio and transcripts never leave it, and nothing is trained on them.
2. lenses
The product's key idea, told before anything is asked of you: every recording is read through lenses — one writes the transcript, one distils the summary, one collects the action items. They're swappable, and you can add more or build your own. Nothing to answer but "good to know".
More in lenses.
3. brain
"Right now I'm all straw and no brains. Mind if I go and get some?"
yapyap sizes up your machine, then recommends one of five presets:
| preset | what it means |
|---|---|
| essential | transcripts and speakers only: the lightest set, for a computer with some years on it |
| featherweight | smallest models — kindest to an older machine |
| balanced | good transcripts without much fan noise |
| performance | the sharpest models, and live words as you talk |
| studio | the biggest models we ship, for a machine with room to spare |
One button takes the recommendation; I'll pick my own opens all five. Either way the honest cost is on screen first — ≈ 3.9 GB to download — and the download then runs in the background. You can start recording straight away; the words fill in once the speech model lands.
Choosing model-by-model, per pipeline stage, lives in settings → models — not in the thread.
4. telemetry
The one real question: may I send anonymous usage counts? Feature counts only — never your audio, transcripts or filenames. It stays off unless you say yes, and "no thanks" is one tap.
5. done
You're set: yapyap is yours to try for 7 days — no key, no email, nothing to enter now. It's a one-time purchase, never a subscription. From here you either record something or head home.
skipping
Skip setup for now is offered on every turn except the first and the last. Skipping doesn't leave you with nothing: it applies the recommended preset and starts the same download. A machine that can record but never transcribe isn't a shortcut, it's a bug.
Older builds read you five sentences to enrol your voice. That step is gone. yapyap now learns which voice is yours the same way it learns everyone else's — you mark a named speaker as this is me. See speakers & voiceprints.