Home » Blog » Home Assistant AI Tutorial: Gemini Welcome Messages and Doorbell Person Descriptions

Home Assistant AI Tutorial: Gemini Welcome Messages and Doorbell Person Descriptions

Home Assistant AI Tutorial: Gemini Welcome Messages and Doorbell Person Descriptions

AI has become a hot topic, and while it brings convenience, it also raises real concernsโ€”especially around privacy and reduced human interaction. Still, there are ways to use AI in a smart home that are lightweight, practical, and fun.

In this tutorial, Iโ€™ll show two Home Assistant AI automations using Google Gemini (Google Generative AI):

  1. A funny AI welcome-home message played via Text-to-Speech
  2. A doorbell notification that describes the person at your door from a camera snapshot

Both use built-in Home Assistant features and a small amount of YAML.


โญโญโญ NOTE: โญโญโญ

This article belongs to a YouTube video. I wrote it before recording the video. You can help me continue doing this by checking out the video, commenting on it, giving it a thumbs up, andย subscribing to my YouTube channel. This will ensure that the video is offered more frequently to new visitors, allowing them to stay informed about the latest Home Assistant tutorials and smart home-related product reviews.

I appreciate your support!

Ed

Click here to watch the video

What You Need

To follow along, youโ€™ll need:

  • Home Assistant
  • A configured person entity with a device tracker (Companion App recommended)
  • A smart speaker (or any media_player entity)
  • A doorbell or camera entity (for the snapshot use case)

Youโ€™ll install:

  • Google Translate Text-to-Speech
  • Google Generative AI (Gemini) integration

Step 1: Install Google Translate Text-to-Speech (TTS)

  1. Go to Settings โ†’ Devices & Services
  2. Click Add Integration
  3. Search Google
  4. Select Google Translate Text-to-Speech
  5. Choose your language and finish

This gives you a reliable TTS engine for your smart speaker.


Step 2: Create a Gemini API Key

Gemini requires an API key:

  1. Open Google AI Studio
  2. Click Get API key
  3. Create a new key and copy it

Step 3: Install Google Generative AI (Gemini) in Home Assistant

  1. Go to Settings โ†’ Devices & Services
  2. Click Add Integration
  3. Search Google
  4. Select Google Generative AI
  5. Paste your API key and finish setup

I need your help!

You will be doing me a huge favor if you subscribe to my channel if you haven’t already. And, you will help me a lot if you also give this video a thumbs up and leave a comment. This way, YouTube will present this video to new people, making the channel grow! In the video description, you will also find information about how you can sponsor me so that I can continue to make these tutorials for you.

Thank you!

Use Case 1: AI Welcome Home (Funny TTS)

Goal

When you arrive home, Home Assistant generates a short welcome message using Gemini and plays it on your speaker.

How it works

  • Trigger: person.ed changes to home
  • Delay: 3 minutes (so youโ€™re actually in the living room)
  • Action 1: Gemini โ€œGenerate Contentโ€ with a funny prompt
  • Action 2: tts.speak using {{ response.text }}

This is a great example of using AI to make your smart home feel more personal without being intrusive.


Use Case 2: AI Doorbell Notification (Describe Person at Door)

Goal

When the doorbell rings, take a snapshot, ask Gemini what it sees, then send the description to your phone.

Workflow

  1. Trigger: doorbell rings
  2. Action: camera.snapshot โ†’ save to /config/www/tmp/snapshot-doorbell.jpg
  3. Delay: 1 second (ensure file exists)
  4. Action: Gemini โ€œGenerate Contentโ€ with image reference + prompt
  5. Action: phone notification with:
    • Message: {{ response.text }}
    • Image: /local/tmp/snapshot-doorbell.jpg

This creates a smart notification like:
โ€œSomeone in a red jacket is holding a package at your doorโ€ฆโ€


Tips for Reliability

  • Use /local/ when sending images in notifications
  • Add a short delay after snapshot creation
  • Device trackers can be slowโ€”donโ€™t expect instant โ€œHomeโ€ detection every time
  • If you want continuous detection and full control, pair this with Frigate or a dedicated NVR

Optional: LLMVision and Local AI (Ollama)

If you want more advanced AI workflows, you can explore:

  • LLMVision (HACS) to support multiple AI providers and richer media analysis
  • Ollama to run AI locally (better privacy, no cloud)

If you want a full tutorial on LLMVision or Ollama in Home Assistant, thatโ€™s a great follow-up topic.


Final Thoughts

These two automations are small, practical examples of using AI in Home Assistant:

  • A fun welcome home
  • A useful doorbell description

Theyโ€™re easy to implement, easy to customize, and they can make your smart home feel more aliveโ€”without turning your house into a โ€œcloud-onlyโ€ AI machine.


If this tutorial saved you time or frustration, consider sponsoring the channel. These guides take many hours to produce.

โญโญโญ

โญโญโญ

Please follow and like us:


Back to all articles