AI
MULTIMODAL
VOICE
WEARABLES
Overview
Live AI is a real-time, multimodal, AI session on Meta smart glasses. Users can ask ongoing questions about what they're seeing. Live AI maintains visual and conversational context as the situation around the user changes.
My role
As a product designer for Live AI, I oversaw the on-glasses interaction across it's early access and general audience launches. My work covered the full conversational arc from entering a session, orienting to context, follow-up questions, and ending the session. I also created interactions for a full landscape of error states, edge cases, and concurrencies.
Live AI
2025 · Meta · Ray-Ban Smart Glasses
1.
SESSION STATUS
How do users know they're in a session?
Standard voice interactions are single turn. Live AI persists. I needed to suggest this notion of a session to users, cue them that AI video was working, and get them interacting quickly. I devised a dynamic opener. It signaled the start of the session, quickly commented on the environment, and asked a brief question to spur engagement.
Starting a session
Reframing an image
2.
RECOVERY PATHS
What if the AI needs more context?
Unlike a camera viewfinder, on glasses users don't have direct visual feedback on what's in frame, so they don't know exactly what the AI can see and read. I designed response templates to help Live AI deliver better answers. These acknowledged visual ambiguity gracefully and dynamically suggested corrective actions to improve responses.
3.
CONCURRENCY
What other things can you do during a session?
Having the glasses mic, speakers, and camera occupied for a Live AI session limits what other features can be used. From incoming calls, outgoing messages, music playback, captures, and more, I developed a comprehensive pattern of in-session interactions to signal guardrails, and give users confidence about what was going on at any given moment.
Concurrent tasks
Add a Title
Add a Title