tessera
All plugins
Voice assistant plugin preview

Voice assistant

Talk to Home Assistant's Assist from the screen: a wake word on the screen itself, the answer on its speaker.

By MaxGramser · v0.2.3

Needs Microphone, Speaker, Media player, PSRAM on the screen.

View on GitHub How to add a plugin
No rating yetRatings come from people who run it, in the Tessera app
1person likes it, some without running it
0homes use it, of those who share that
How these numbers are counted

Everything here is shared by choice from the Tessera app, one per installation of the app, without an account. A like can come from anyone, also from someone who does not run the plugin. A rating comes from someone who has run it on a screen for a few days, and shows from 3 ratings. Homes are the installations that agreed to be counted and said in the last 30 days that they run it.

These are not verified installations or a quality score. Reinstalls and lost keys can affect them, and one home network adds at most two of each to a plugin.

What the app shares

Talk to Home Assistant's Assist from the screen. The screen listens for a wake phrase on its own, and only then sends what you say to Home Assistant; the answer comes from the screen's speaker.

Speech to text, the conversation and text to speech are whatever your Home Assistant uses for Assist: its own conversation agent with Whisper and Piper, Home Assistant Cloud, or an AI such as OpenAI or Claude. The screen is a voice satellite in Home Assistant, in the area you give it.

What you need

  • A screen with a microphone and a speaker: a board that brings them (the reTerminal D1001) or a plugin that does.
  • An Assist pipeline in Home Assistant (Settings, Voice assistants) with speech to text and text to speech.

Set up

  1. In Tessera, open Plugins, choose Voice assistant and tick your screens. Each screen builds once with it.
  2. In Home Assistant, open the screen's device (Settings, Devices) and choose its Assist pipeline under Assist pipeline, if it is not the preferred one.
  3. Put the tile Assistant on a page: it says what the assistant does, and a tap asks a question without the wake phrase, or stops one.
  4. Optional: add Assistant to a page's top bar (Top bar, Add, From plugins). It shows while the assistant listens, thinks or answers.

On the screen

Hold the top bar, then Settings, Plugins, Voice assistant:

  • Wake word: whether the screen listens for its wake phrase. Off, a tap on the tile still asks.
  • Wake phrase: Okay Nabu, Hey Jarvis or Hey Mycroft.
  • Wake sound: a short sound when the screen starts to listen, after the wake phrase or a tap.
  • Thinking sound: a soft sound, again and again, while Home Assistant works out the answer.
  • Ask now: asks a question at once.

The settings are entities of the screen in Home Assistant too ("Wake word", "Wake phrase", "Wake sound", "Thinking sound"), and the editor shows them in the plugin's details on the screen's Plugins tab. The screen's own Microphone switch, on a board that has one, silences it for the assistant as well.

The sounds

They are two files in components/voice_assist/sounds/: wake.wav and thinking.wav (the screen plays it again as soon as it ends, so its length sets the rhythm). Each is a WAV file of 16-bit samples, mono, at 48 kHz like the screen's speaker. A plugin of your own made from this one can swap them; the build says so when a file is not such a WAV.

Privacy

The wake phrase is recognised on the screen itself; nothing leaves the screen until it is heard or you tap the tile. After that the question goes to your Home Assistant and on to whatever speech and conversation services its pipeline uses.