The problem is not too few devices, it is too many steps

The living room is usually the densest room in the house. Ceiling light, cove lighting, a floor lamp, blinds, air conditioning, a television, speakers. Over one evening you switch between two or three different arrangements. To watch a film you kill the ceiling light, dim the cove lighting, lower the blinds, drop the temperature a little, then switch the television to the right input. Pick up the phone, swipe around, and a dozen taps have gone by.

No single tap is hard. The count is what kills it. Once the count is high enough, people cut corners, and everyone ends up walking back to the wall switch again. All that careful setup goes unused.

A spoken command exists to fix exactly that: it compresses those dozen taps into one sentence.

The word "voice" here may not mean what you think

Home Assistant ships with a conversation agent, known in the system as conversation. It does one job: understand a sentence and decide what to do about it. Here is the part that matters. What that layer actually consumes is text. Typing into the chat box in the phone app and speaking to a voice satellite arrive in front of it as the same thing.

So you can get the whole set-up working by typing, confirm every phrase behaves, and add a voice satellite later if you feel like it. Nothing extra gets installed on the hub, and no local speech engine is involved. That is the single most important point in this article: the barrier to entry is zero.

Naming a command so you still remember it

Most people set up a spoken command once and never use it again, for a very plain reason: they cannot recall the name they picked. How well you name things decides whether this lasts a week or three years.

First principle: use words you already say. Settling into the sofa, what comes out of your mouth is "movie time", not "execute scene three". Second principle: name the intent, not the device. What you want is the mood of a film night, not the third lamp at twenty percent; how the hardware complies is the system's problem. Third principle: give each scene one primary name, then two or three alternate wordings, so it still lands when everyone in the house phrases it slightly differently.

What happens after the sentence leaves your mouth

From the end of your sentence to a fully rearranged room takes a little over a second. Several layers hand off inside that second.

1. Second 0

   You say it

   Into the chat box in the phone app, or typed straight in. If a voice satellite is on the shelf, speaking to it works the same. All three routes look identical to the system.

2. 0.3 sec

   The conversation agent recognises the phrase

   It compares what you said against the wordings you registered and picks the closest one. Register "movie time" and "movie" and something like "let us watch a movie" usually still lands.

3. 0.5 sec

   The blueprint maps the phrase to actions

   Once recognised, it hands over to a rule built on voice_trigger_things. Its only job: hear this phrase, carry out that set of actions.

4. 0.8 sec

   The scene fires first

   The living room lamps land together rather than lighting up one by one. The blinds start travelling at the same instant.

5. 1.5 sec

   The rest catches up

   Cooling goes to 25, the television switches to the input you wanted. The blinds need another dozen seconds to finish, by which time you are already seated.

6. A minute later

   A record arrives on your phone

   A push message lists what ran and whether anything failed to respond. Nothing is announced out loud; the note simply sits in your notification list.

Four things it quietly takes care of

Before and after

What you need on hand

Home Assistant files every device in the house as an "entity". Brand and model do not matter; the category does. Have the ones below and you can start.

How the whole thing works

The path has three segments: a sentence comes in, something recognises which sentence it is, and a set of actions goes out.

Use the rule that already exists

The hub ships with a tested set of automation blueprints. Think of a blueprint as a fill-in-the-blanks form: the boxes are drawn for you, you only supply the content. This situation uses voice_trigger_things, which exists precisely to sit behind the conversation agent.

Fill those five boxes and the phrase is alive. For a second or third phrase, build another rule the same way; they do not interfere with each other.

A physical button can call the same scene

Some moments you simply do not want to speak, the middle of the night for instance. Use switch_trigger_things to wire a wireless button on the wall or the side table to the same scene, and pressing it does exactly what the sentence does. Keep both and use whichever is closer. Since they call the same scene, changing it later means editing one place.

Different homes, different phrases

A few common situations and how to adjust.

You do not have to set this up yourself

None of the above needs to be clicked in box by box. The WOOWTECH smart hub runs HA-MCP, so an AI can connect straight into your system, do the work, then test it itself and confirm it really runs before reporting back.

Prompts you can copy straight out

Every block below can be copied whole with the button on the right. Pick the one closest to what you need; there is no need to use them all.

Set up one spoken command for my living room.
When I say "movie time", dim the living room lamps and lower the blinds.
It has to work by typing as well as by voice.

Build me a set of spoken living room scenes, three phrases in total.
"Movie time": lamps to 20%, blinds fully down, cooling at 25, TV on the playback input, one small lamp left on in the hall.
"Heading to bed": all living room lamps and the TV off, blinds down, cooling to 27.
"Guests are here": lamps to 90%, blinds open, cooling at 24.
All three must work by typing or through a voice satellite, and do not lock the wording down; "movie" and "let us watch a movie" should land too.
Send me one push message after each run listing which devices moved. Do not announce anything out loud.
When you are done, test it yourself, confirm it really works, then tell me which devices you used.

My living and dining areas are open plan and their lamps are wired together.
For "movie time", dim only the lamps over the sofa and leave the dining side as it is.
Also add a phrase "dinner is ready" that puts dining lamps at 100% and holds living room lamps at 40%.
The two phrases must not interfere; I may use them one after the other.

"Movie time" is dimming the lamps too far now, make it 35%.
Add one more wording to that phrase, "movie mode", because I sometimes say it that way.
Leave everything else alone.

A few of the things I say get no response, and my wording may not match yours, or the phrase may never have been set up at all.
Leave the settings alone for now.
Write out every wording the living room understands right now, grouped by rule, with all the alternate phrasings listed under each one.
I want to hold that list up against the way I actually talk.
Once I have read it I will mark the ones I use, and we can fill in what is missing together.

Common questions