Per-voice avatars: every Grok voice gets its own HUD-style character on the dashboard

[noisy]

Idea

Each Grok voice (currently 26: altair…zenith) gets its own avatar shown on the dashboard — a character drawn in the HUD style, consistent with the rest of the interface. At a glance you see WHO is speaking, and the voices gain identity.

Materials

• A ready-to-use Gemini prompt for a 26-avatar sprite sheet in the HUD style (colors and grid derived from dashboard/src/styles/hud.css) lives in issues-drafts/gemini-avatar-sprite-prompt.md.
• 6×5 grid, 341×341 px tiles — trivial to slice programmatically; the order matches the API's alphabetical voice list, so index → voice_id is deterministic.

Implementation notes

• The voice list comes from the xAI API (tts.py: list_voices) — key avatars by voice_id, with a generic-silhouette fallback for voices the API adds later.
• The dashboard knows the current voice per conversation — show the avatar on the tab / utterance card.
• Pairs naturally with per-agent colors and the character-per-tab layout ideas — all of these are agent-identity signals.
• Follow-up (separate issue): lip-sync animation of the avatar's mouth while TTS plays.

🤖 Generated with Claude Code

[noisy]

example file