Build Your Space AI Agent

Preset: Face Recognition {#face_recognition}

Add face recognition to your Xiaozhi, letting it recognize family and friends.

What you'll get:

• Automatic greeting when recognized face appears
• Enroll and manage faces from the web panel (up to 20 people)
• Voice control to enable, query, and delete faces

Requirements: WiFi network · Xiaozhi App for device binding

Step 1: Flash Xiaozhi Firmware {#face_esp32 type=esp32_usb required=true config=devices/watcher_esp32.yaml}

Write the voice assistant program to the Watcher to enable voice interaction.

Wiring

1. Connect Watcher to computer via USB-C cable
2. Select the serial port above (choose one starting with wchusbserial)
3. Click the Flash button

Troubleshooting

Step 2: Flash Face Recognition Firmware {#face_himax type=himax_usb required=true config=devices/watcher_himax.yaml}

Write the face recognition program to the Watcher's AI chip.

Wiring

1. Ensure Watcher is connected to computer
2. Select the serial port above (choose one starting with usbmodem)
3. Click the Flash button
4. After clicking Flash, press the reset button on the device to enter flash mode

Troubleshooting

Step 3: Configure Xiaozhi {#face_configure type=manual required=false}

Connect the Watcher to WiFi and bind it to your account using the mobile app.

Connect to WiFi

Device will prompt for network setup on first boot. Follow voice instructions to connect to WiFi.

Bind Xiaozhi Account

1. Open Xiaozhi App
2. Scan the QR code displayed on device
3. Complete the binding process

Test Voice

Wake up the device by saying "Xiaozhi Xiaozhi" to test voice interaction. If it responds normally, the setup is successful.

Troubleshooting

Step 4: Face Database Management {#face_enroll type=serial_camera config=devices/face_enroll.yaml required=false}

Manage the face recognition database through the app interface.

How to Use

1. Click Connect to start the camera preview
2. You'll see a live feed with face detection boxes
3. Use the Face Database panel below to manage enrolled faces

Register a New Face

1. Click Register in the Face Database panel
2. Enter a name for the person
3. Click Start Capture — face the camera with good lighting
4. Wait for the capture to complete (5 seconds)
5. The new face will appear in the table

Troubleshooting

Deployment Complete

Face recognition is ready! Let's try it out.

Step 1: Enroll faces from the panel above

Use the Face Database panel to enroll family or colleagues (see "Register a New Face" above).

Step 2: Enable face recognition by voice

Unplug the USB cable, wake up the device by saying "Xiaozhi Xiaozhi", then say:

"Turn on face recognition"

Once confirmed, Watcher will automatically scan for faces while idle.

Step 3: Experience automatic recognition

Walk in front of the Watcher — it will recognize you and greet you by name!

Voice Commands

Preset: Display Cast {#display_cast}

Cast Xiaozhi conversations to TV or large display, ideal for exhibition halls, meeting rooms and multi-person scenarios.

What you'll get:

• Real-time conversation display on big screen
• Fullscreen mode for presentations
• mDNS auto-discovery - connect by voice command
• Narrate mode - AI controls background images for storytelling & guided tours

Requirements: All devices on same network

Step 1: Flash Watcher Firmware {#display_watcher type=esp32_usb required=true config=devices/display_watcher.yaml}

Write the voice assistant program to the Watcher for display casting.

Wiring

1. Connect Watcher to your computer using USB-C cable
2. Select the serial port above
3. If not detected, try a different USB port or cable

Troubleshooting

Step 2: Deploy Display Service {#display_service type=docker_deploy required=true config=devices/display_local.yaml}

Start the display service that shows conversations on your screen.

Target: Local Deployment {#display_service_local type=local config=devices/display_local.yaml}

Deploy the display service on your local computer.

Wiring

1. Ensure Docker is installed and running
2. Set a display name (e.g. "Living Room Display") for mDNS discovery
3. Click Deploy button to start services

Troubleshooting

Target: Remote Deployment {#display_service_remote type=remote config=devices/recomputer.yaml default=true}

Deploy the display service to reComputer R1100.

Wiring

1. Connect reComputer to network and HDMI display
2. Enter IP address and SSH credentials
3. Set a display name (e.g. "Meeting Room Display") for mDNS discovery
4. Click Deploy to install on remote device

Troubleshooting

Deployment Complete

Display cast is ready!

Test it:

1. Open http://<device-ip>:8765 on display browser
2. Press F for fullscreen mode
3. Say "Cast to [Display Name]" to start

Voice commands: "Start casting", "Stop casting", "Cast status"

Narrate Mode (New): The display now supports a narrate mode where AI can control background images — ideal for presentations, storytelling, and guided tours.

1. Click the gear icon (top-left) to open the config panel
2. Toggle "Enable Narrate Mode"
3. Enter your Xiaozhi WebSocket MCP URL to enable AI-driven image switching
4. Add trigger rules: keyword + image URL pairs for automatic background changes
5. Press N to toggle narrate mode, click the PiP window to resize it