[Feature]: Clearer capture-permission and privacy copy (Voice)

[rosspeili]

One-line summary

Review and improve in-app + docs copy for screen/audio capture permissions so users know what is captured and that it stays local.

Problem / motivation

Roadmap item. Device/window capture can trigger OS prompts. Vague status strings (starting, error) and thin privacy wording increase distrust for a tool that listens to system audio next to agents.

Proposed solution

• Voice panel: short explainer under Device output / Pick window.
• Error states: actionable text (permission denied → Settings deep-link hints on Windows).
• Docs: Audio sources + Privacy blurb in README already — expand “what never leaves the machine.”
• Optional one-time first-run note when enabling capture.

Alternatives considered

• Leave as-is — ships, but weak for security-minded users.

Primary surface

Voice / lip sync

Constraints you accept

• Should stay usable offline / local-first
• Browser-only mode may remain limited vs Electron

Mock / sketch / reference (optional)

Inline note: “AVATAR analyses audio levels locally. Nothing is uploaded.”