Skip to content

Open source · Apache 2.0

A voice agent that lives on your hardware.

Talk to it and it answers out loud, in a voice and character you shape. It remembers you, acts through the skills you connect, and stops to listen when you talk over it. No cloud, no accounts.

Domia, the local voice agent character
DomiaHallway

“Who are you, and what do you do?”

I'm Domia, the AI companion built to live with you, helping with the people, conversations, and music in this house.

language modelfirst audio in 2.6 s

See it work

One evening at home, as one example: a light, someone it remembers, a guest asking for the wifi and a good night. Every reply and every timing in the video is what Domia produced as it ran.

Hear it answer

Real turns from the demo setup. Press play: you hear the question, the wait and the reply.

  • MarloweCinema

    “My name is Laura, and I can't stand horror films.”

    Laura, horror's not for you, I agree.

    language modelfirst audio in 1.6 s
  • MarloweCinema

    “Marlowe, what do you remember about me?”

    Laura, you mentioned you can't stand horror films, and we just talked about it.

    rememberedfirst audio in 2.2 s
  • AtlasGuest room

    “What's the guest wifi password?”

    sunflower2026

    knowledge basefirst audio in 1.1 s

Make your Domia a person

A Domia is someone: a name, a voice, a temperament and a way of speaking. Here are five from the demo setup answering the same question. Press play and hear how different they are.

Ask them:
  • AtlasCalm host

    It was quiet so far.

    0:02
    first audio in 1.1 s
  • LunaEmpathetic storyteller

    It was a bit sleepy, but I love hearing about your day.

    0:03
    first audio in 948 ms
  • MarloweAnalytical film buff

    It was a dark one, full of twists.

    0:02
    first audio in 1.3 s
  • TorqueCautious technician

    Not bad.

    0:01
    first audio in 739 ms
  • SousPlayful chef

    It's been a revelation, full of unexpected flavors and aromas that have left me breathless.

    0:05
    first audio in 1.2 s
  • Yours

    In the console you pick the look and the voice, set the mood on a radar and write the character in your own words.

Where does your voice go?

With Domia the whole turn runs on a machine you own, so audio, transcripts and memory never reach a third party.

The same three questions, answered for both

Where is your audio processed?

A cloud assistant
On the vendor's servers
Domia
On a machine you own

Who keeps a copy?

A cloud assistant
The vendor, with your account and logs
Domia
Only your node, where you can read and erase it

What happens when the internet drops?

A cloud assistant
It stops answering
Domia
It keeps working, and so does everything on your network; only cloud services stop

Where it fits

  • At home
    Lights, music, timers and questions, answered by someone who knows the place.
  • Hospitality
    Each room answers guests from the notes you wrote: wifi, rules, local picks.
  • Attractions and guides
    A character at a park or a guide in a museum, in character, with its own voice and notes.
See the use cases

A console for every node

Every node you run, in one place: who is online, what each one heard and said, and what it remembers. Apply a template or build a configuration from scratch, and edit routines, satellites and memories from the browser.

The console overview: fleet counters and a map of the Domias on the local mesh

Start with one machine

Follow the guide, give it a configuration from the console, and say the wake word.

Source on GitHub (opens in a new tab)
Prefer a second opinion? Ask your AI about Domia

Open the assistant you already use with a short prompt that points it at the sources.