Back to live

Settings

Speaker identification

Voice training set

Not trained yet. Record 3–4 varied clips for reliable matching.

Read naturally: “Thanks for having me — I'm happy to walk you through my background and what I've been working on recently.”

Setting a number stops one remote speaker being split into several — extra voices are folded into the closest known speaker.

Layout

Left/right dock as a narrow sidebar; top/bottom spread into a wide bar. Keep this window beside your meeting app — browsers can't paint a true transparent overlay.

Answer pacing

AI-suggested length

Target adapts to each question instead of a fixed number.

Drift guard

Nudge me when my answer wanders off the question.

Session type