Voice Control

Wrist Assistant works with Home Assistant Assist. Speak to your watch to control devices or ask questions. Assist is very literal — you need to say the exact entity names as they appear in Home Assistant for it to understand you.

Examples:

• “Turn off the living room lights”
• “Set the thermostat to 72”
• “Is the garage door open?”
• “Lock all doors”

Trigger Assist from a quick menu slot or a dedicated Assist tile placed on any page.

TTS Quick Menu

Add a text-to-speech action to any tile’s quick menu. Long-press a tile, slide to the Speak Message slot, and your message plays on the speaker you’ve chosen. Set your TTS engine and default speakers in Settings → Voice & Speech.

Examples: “Dinner’s ready” on the kitchen speaker, “Time for bed” on the kids’ room speaker.

Voice Phrases (Pick from List)

Save up to 8 reusable TTS phrases so you can pick from a list instead of dictating each time. Speak Message (dictate a one-off) and Pick from List (your saved-phrase library) are separate quick-menu slots you add independently — long-press a tile, slide to the Pick from List slot, and the phrase library appears. Manage the phrases in Settings → Voice Defaults → Pick from List - TTS.

Each phrase can override the default TTS engine and target speakers, so “Dinner’s ready” can hit the kitchen speaker while “Time for bed” hits the kids’ rooms.

Voice Broadcast

Record a live voice message directly from your watch and play it on your Home Assistant speakers. Instead of picking a pre-set phrase, just talk — your voice gets broadcast to whichever speakers you choose.

Add a Broadcast slot to any tile’s quick menu. Long-press, slide to the broadcast icon, speak your message, and it plays throughout the house.

Assist Satellite & Voice Entities

Home Assistant voice entities open their own control screens when you add them as tiles:

• Assist Satellite (assist_satellite.*) — A control view with Announce, Chat, and Tools tabs: make an announcement (with an optional pre-announce tone), start a conversation, ask a question with expected answers, open the Notify composer, and jump to its linked conversation / tts / stt / wake_word entities.
• Conversation / TTS / STT / Wake Word (conversation.*, tts.*, stt.*, wake_word.*) — Each opens a focused control view: send or dictate a command to a conversation agent, speak or dictate a message through a tts.* entity, or dictate to Assist.

Siri

You can also control devices through Siri without opening the app. See Siri Shortcuts for details on what voice commands are available.

Output Modes

Where Assist replies play is configured per slot in the quick menu / radial editor:

• Silent Reply on Watch — Runs Assist and shows the reply on the watch without speaking it
• Speak Reply on Watch — Speaks the reply through the watch’s built-in speaker
• Speak Reply on Speakers — Speaks through Home Assistant speakers — either this tile’s configured speakers or the global defaults
• Choose on Watch — Lets you pick a speaker from this tile’s list each time

TTS actions (Speak Message and Pick from List) route to Apple Watch, Choose Speakers, or Choose on Watch. Broadcast routes to Specific Speakers or Choose on Watch only — there’s no Apple Watch option, since live audio plays on your Home Assistant speakers, not the watch.

When speaking through Home Assistant speakers, you can also control how the speaker’s volume is handled — keep the current volume, set a level for the message and restore afterwards, or set a level and keep it.

Voice Input

Enable Start Listening Immediately on a slot to begin dictation as soon as the action opens — useful for a hands-free flow. When Assist needs to disambiguate (“Which light?”), the watch automatically prompts for another utterance instead of dropping you back to the tile.

Conversation Agent

Assist tiles, quick menu Assist slots, and radial Assist slots all let you pick a specific conversation agent (any conversation.* entity — typically an LLM-backed assistant like an OpenAI / Anthropic / local-model integration). When set, Assist routes the request through that agent instead of Home Assistant’s default Assist pipeline. Useful when you want one tile to talk to your “house brain” LLM and another to use HA’s built-in intents.

• Per-tile / per-slot — Pick the agent on the tile’s Action tab (Assist tiles), or on the Assist slot inside the radial / quick-action editor. Leave it blank to fall back to the global default.
• Global default — Set under Settings → Voice & Speech → Assistant. Applies everywhere no tile-level agent is set; reset returns to Home Assistant’s normal Assist route.

Setup

In Settings → Voice & Speech, pick your TTS engine (Google Translate, Piper, etc.), default speakers, and (optionally) a default conversation agent. The app will try to seed a recommended engine automatically if one is available. You can override the defaults per TTS quick menu slot or per saved phrase if you want different speakers in different rooms.