Betty is an AI voice companion for long-haul truck drivers. She proactively calls drivers
when safety systems detect fatigue or risky driving, having real conversations that save lives.
Built with Google Gemini Live API · Real-time bidirectional voice · Cross-call memory
📞
Proactive Calling
Betty calls the driver — not the other way around. When a fatigue camera detects drowsy eyes or a telematics system flags erratic driving, Betty picks up the phone.
🧠
Cross-Call Memory
Betty remembers previous conversations within a shift. If she called about fatigue an hour ago, she'll reference it: "Last time we spoke you were looking tired..."
🛑
Smart Escalation
If a driver refuses to pull over, Betty escalates to the fleet manager with context — what happened, what she tried, and why she's concerned.
🖼️
Visual Cards
Betty sends visual cards to the driver's dashboard — rest stop recommendations with AI-generated backgrounds, shift wellness summaries, and incident reports.
How this demo works: Each page demonstrates a key feature.
You'll trigger real AI conversations and see Betty respond in real-time.
All calls are simulated — an AI driver persona talks to Betty so no microphone is needed.
Step 1 of 5
Fatigue Detection
When an in-cab camera detects signs of fatigue — droopy eyes, yawning, or head nodding —
Betty receives the video feed and calls the driver to check on them.
What happens: A fatigue camera event fires. Betty receives video frames showing
the driver's fatigue behaviour, then calls to have a caring conversation about pulling over for a rest.
In-Cab Fatigue Camera Feed
Sample fatigue camera footage — droopy eyes detected
Frames Sent to Betty
These JPEG frames are extracted at 1fps and sent to Betty's Gemini session via
send_realtime_input(video=...)
so she can see what the camera detected.
Extracting frames...
Try it — trigger a fatigue event
Press the button to trigger a fatigue event
00:00
Conversation will appear here...
Complete the call to continue
Step 2 of 5
Smart Escalation
When a driver shows dangerous behaviour and Betty can't convince them to stop,
she escalates to the fleet manager with a detailed incident report.
What happens: A high-severity head-nod (microsleep) event fires. The simulated
driver is stubborn and refuses to pull over. Betty escalates and generates an incident report card.
Try it — trigger a microsleep with a stubborn driver
Press the button to trigger a microsleep event
00:00
Conversation will appear here...
Step 3 of 5
Cross-Call Memory
Betty remembers what happened in previous calls within a driver's shift.
If she called about fatigue earlier, she'll reference it in the next conversation.
What happens: First, a check-in call establishes context. Then a second fatigue
event fires — listen for Betty referencing the earlier conversation.
Memory is encrypted with AES-256-GCM and auto-expires after 14 hours.
Try it — two calls, same shift
Start with Call 1 — a friendly check-in
00:00
Conversation will appear here...
Step 4 of 5
Shift Summary
At the end of a shift, Betty generates visual summary cards — wellness reports
with shift timelines, rest stop recommendations with AI-generated scenic backgrounds,
and incident reports for the fleet manager.
Try it — generate visual cards
Rest stop cards use Gemini Flash + Google Search grounding to describe the real
location, then Imagen 4 generates a scenic background photograph.
Step 5 of 5
How It Works
Betty is built entirely on Google's AI stack — Gemini Live API for voice,
Gemini Flash for grounding, and Imagen 4 for visual card generation.
Voice Call Flow
Safety Trigger→Risk Scoring→Gemini Live API↔WebSocket Bridge↔Driver (Browser/Phone)