It stopped working: which permissions should I check?

Open System Settings, then Privacy & Security, and make sure Writhere is enabled under Microphone, Accessibility and Input Monitoring (plus Screen Recording if you transcribe calls). Then relaunch the app: macOS applies some of these only to a fresh start. You can check them anytime from the Autorizzazioni section in the app dashboard.

What are the system requirements?

A Mac with macOS 13 or later. Apple Silicon (M1 or newer) is recommended: everything runs on-device, and the AI features are dramatically faster there. Intel Macs work too, with longer processing times for call transcription and summaries.

Why does it ask for the Microphone?

To hear your voice while you dictate. The audio is processed on your Mac and never leaves it.

Why does it ask for Accessibility?

To type the dictated text straight into your apps and to catch the shortcut in any app. Without it, Writhere would be stuck in its own window.

Why does it ask for Screen Recording?

Only to hear the other side of a call you want transcribed: macOS gates system audio behind this permission. It is optional, nothing is filmed, and no audio ever leaves your Mac.

€49 forever — what's the catch?

There isn't one. The AI runs on your device, so there's no ongoing cost on our side to pass on to you. One payment, yours to keep.

Is my voice really private?

Yes. Everything is processed locally. Your audio and text never leave your machine — there's no server to send them to.

What happens when I switch computers?

Your license covers every device you own. Install Writhere on the new one and you're set — no account, nothing to migrate from the cloud.

Yes — improvements land as free updates, so the app you own keeps getting better at no extra cost.

July 23, 2026

A Dictation App That Uses Zero Memory When You're Not Talking

Open Activity Monitor while a dictation app just sits in the menu bar, not actively listening, and you'll sometimes find it quietly holding hundreds of MB of RAM anyway. That's not a requirement of doing speech recognition well, it's a design choice, and not always a good one.

Quick answer

A speech model only needs to be in memory while it's being used. Loading it on demand and unloading it after a short idle window means the app costs you something while you dictate, and close to nothing the rest of the time.

Why "always loaded" became the default

Keeping a model warm avoids the load time the next time it's needed, so some apps just leave it running permanently. That trades a few seconds of occasional startup delay for a constant, all-day memory and battery cost, on every machine, whether you dictate once an hour or once a day.

What idle-aware memory management actually looks like

— Load on first use, not at launch. The model comes into memory the moment you actually start dictating.
— Unload after a short idle window. A few minutes of inactivity is enough to release the memory back to the rest of the system.
— Scale down on machines that need it. A Mac with less RAM can use a smaller model automatically, rather than one-size-fits-all regardless of hardware.

Why this matters beyond a number in Activity Monitor

On a machine already running a browser with thirty tabs, a chat app, and a design tool, an extra permanently-resident 800MB is the difference between a Mac that feels responsive and one that starts swapping. A background utility shouldn't be the reason the rest of your work slows down.

The bottom line

The right amount of memory for a dictation app to use while you're not talking is close to zero. Anything more than that is a tradeoff someone chose to make, not a limitation of the technology.