* paying a monthly fee for the privilege of using my own API key,

* hauling an 800 MB local model, or

* copy-paste gymnastics from a floating window.

So I built VTS (Voice Typing Studio).

What it is:

* 5 MB notarised menu-bar app. Press a toggle hot-key → talk → text lands wherever the cursor is.

* Bring-your-own OpenAI / Groq key. Nothing touches my servers.

* Universal binary – runs on Intel and Apple silicon.

* MIT-licensed; core stays free/open. If future features justify it, a pro tier may appear.

Repo, DMG, and 8-second demo video: https://github.com/j05u3/VTS

Why it’s different

Previous HN tools solve pieces of the problem, but not all at once:

* superwhisper – fully offline, but big model and Apple-silicon only.

* MacWhisper – accurate, yet still a separate window/copy-paste flow.

VTS aims for “menu-bar switch and forget.”

Would love feedback on onboarding, features or anything that breaks.