A Face You Can Film

The speaking face is a wireframe core and a few teal rings on a black square. It measured 27,928 kB, and the rings were moving outward.

Savindu Thathsara, savithathsara.me. This is a Hearthlen note from the same ThinkPad.

Motion is the sentence

A face that only changes color is hard to read in a recording. This one has a direction for each state. Listening pulls the rings inward, toward a core that stays still. Thinking holds the rings and turns the core, so the meridians sweep. Speaking sends the rings outward, and the words being said sit underneath. A person, or another model, watching a video of that window can tell the three apart without the source.

It is still the same window as before: built once, shown only while Hearthlen is in the conversation, hidden by a click or when the sentence ends. The desktop path that typed an address in the previous note is unchanged. The face is just the thing you see while that happens.

There is no image and no clip to decode. Twenty-eight points make a ring. A handful of ellipses make the core. The redraw waits 50 milliseconds. The process RSS while Speaking was on screen was 27,928 kB.

---

Savindu Thathsara builds practical AI and automation in London. This series is the public log of an offline AI operating system he is writing and learning at the same time.

• Site: https://savithathsara.me
• Email: [e-mail]
• DEV: https://dev.to/savinduthathsara