Camera Snapshot, Blue Iris, AI, Deepstack, Input Boolean, Notification

I use Blue Iris with VorlonCD AI Tool & Deepstack to detect motion. I have the VorlonCD AI Tool send an mqtt message when relevant motion is detected. I then use node-red to receive the message and turn on an input_boolean, then after 5 seconds the boolean is turned off. This blueprint will send a notification with image to your device. I can’t take full credit for this blueprint. I just adjusted an existing blueprint posted on this forum to fit my needs with an input_boolean.

All my cameras are in Home Assistant via the HACS Blue-Iris integration.

Here’s the original that I modified.![BI_AI-Blueprint

Get started

Click the badge to import this Blueprint: (needs Home Assistant Core 2021.3 or higher)

Open your Home Assistant instance and show the blueprint import dialog with a specific blueprint pre-filled.

Or import this Blueprint by using the forum topic URL:

  name: Send Camera Snapshot on Input Boolean state change
  description: 'Sends a camera snapshot when on Input Boolean state change.

  domain: automation
      name: Input Boolean
      description: The Input Boolean that triggers the snapshot creation
          domain: input_boolean
      name: Camera
      description: The camera that creates the snapshot
          domain: camera
      name: Device to notify
      description: Device needs to run the official Home Assistant app to receive
          integration: mobile_app
      name: Is it an iOS device?
      description: Toggle if your selected device runs iOS, default is Android
        boolean: {}
      default: false
      name: Notification title (Optional)
      description: 'Default: "Motion detected!"'
      default: Motion detected!
      name: Notification message (Optional)
      description: 'Default: "{{ input_boolean_name }} detected movement!"'
      default: '{{ input_boolean_name }} detected movement!'
      name: Delay (Optional)
      description: Wait before creating camera snapshot
      default: ''
          min: 0.0
          max: 60.0
          unit_of_measurement: seconds
          mode: slider
          step: 1.0
  platform: state
  entity_id: !input 'input_boolean'
  from: 'off'
  to: 'on'
  input_boolean: !input 'input_boolean'
  input_boolean_name: '{{ states[input_boolean].name }}'
  camera: !input 'camera'
  notify_device: !input 'notify_device'
  is_ios: !input 'is_ios'
  notification_title: !input 'notification_title'
  notification_message: !input 'notification_message'
  delay: !input 'delay'
  snapshot_create_file_path: /config/www/tmp/snapshot_{{ states[camera].object_id
  snapshot_access_file_path: '{{ snapshot_create_file_path | replace(''/config/www'',''/local'')
- delay: '{{ delay }}'
- service: camera.snapshot
  entity_id: !input 'camera'
    filename: '{{ snapshot_create_file_path }}'
- device_id: !input 'notify_device'
  domain: mobile_app
  type: notify
  title: '{{ notification_title }}'
  message: '{{ notification_message }}'
  data: '{% set android_data = {"image": "%s"} | format(snapshot_access_file_path)
    %} {% set ios_data = {"attachment": {"url": "%s", "content_type": "JPEG"}} | format(snapshot_access_file_path)
    %} {{ ios_data if is_ios else android_data }}'


i saw your message in blueiris forum. there you mentioned nodered and mqtt, can you please explain how you using them?

I don’t understand what’s the difference between this modified one to the original one…
Isn’t it just takes the snapshot on motion from BI? it doesn’t matter what kind of AI runs in the background, so I’m confused what is the “improvement” here.

First time trying a Blueprint. I’d like to notify two devices (both iPhone). I tried making a Group (group.mobile_app) but the input selector did not recognized it. What is the recommended technic?

What did you use as input boolean? I am presently using mqtt directly from hassio and node-red to store snapshots. This looks like a simpler way to do this.

I;m still using this. im using the AI tool from here, that works fantastic, together with my hosted AI detector on proxmox, while deepstack is kinda poop. i use open AI.

But i also tried with google generated AI, letting the detections store on my NAS and queue it in homeassisant and send it to google, as this is free of charge when adding creditcard, it does work great but not kinda like i want it.

google ai prompt

Describe the image from my front gate camera if it shows a person, animal, or car (exclude the **YOURCAR** on the driveway). For cars, include the brand, license plate number, and country. For people or animals, describe their actions. If a person has text on their shirt, read it. Only identify gender if certain. Ignore stationary objects and buildings.

Well it did execute really well, but where my camera is positioned its not that great.
However my other camera facing my land it works perfect, it indeed reads the tshirt text.

Using flash version of google.