AI
MULTIMODAL
VOICE
WEARABLES
Overview
Tailored to everyday athletes, Autocapture is an AI-powered capture feature on Meta smart glasses. It syncs with Garmin to automatically capture clips during a workout. After the workout, it packages the clips into an easily shareable reel.
My role
I was the conversation design lead for the on-frame Autocapture experience. Responsible for the full voice interaction layer, I crafted the invocation interaction, overall session structure, concurrency handling, as well as errors and edge cases during the experience.
Autocapture
2026 · Meta · Oakley Smart Glasses
1.
EDUCATION
How do you know what's being captured?
To supplement in-app user education, I designed a tapered session start confirmation. It quickly lets users know a session has begun, and briefly outlines what will be captured for that specific activity. It tapers significantly in subsequent uses to get out of the way. I also created feature help interactions so users can directly ask what's being captured.
Starting a session
Designing for active users
2.
ENVIRONMENT
How do you design for a user in motion?
Athletes can't easily look at a screens or issue voice commands mid-activity. I focused on having brief interactions with minimal back-and-forth once a session was active. When triggered, captures are signaled by earcons and not dialogue. Users can easily pause and resume sessions with both voice and hardware controls.
3.
COMPLEXITY
How do you abstract complexity away from users?
With a Garmin device added to an already intricate dance between glasses, phone, and server, I had to keep things simple but still keep users informed. I designed edge and error state interactions that used everyday language and kept the experience moving forward positively.
For example:
Users can start a session without an active Garmin activity. They'd get a response indicating that Autocapture was active but waiting for Garmin to connect before it captures anything.
Managing a network of devices
Add a Title
Add a Title