Home Assistant’s voice stack used to be the platform’s most-asked-about weakness. The 2024 voice satellite work and the 2025 Year of the Voice push fixed most of that, and the XDA piece on Home Assistant’s AI tool captured the current state: it works, it’s local, and the onboarding is still rougher than Alexa’s out-of-box experience. The phone side of voice control is where most owners still feel friction — what app do you actually open to talk to your house, and what runs in the background to listen?
We tested seven apps and tools for Home Assistant voice control on Android, covering the official Companion app, third-party voice satellites, the cloud-bridged option, and the fully offline picks. Each entry is rated for setup difficulty, response latency, language coverage, and how it handles the “hey, turn off the lights” flow that should just work without an essay of documentation.
What to look for in a Home Assistant voice app
Six things matter when voice is the interface:
- Wake-word reliability that triggers on your phrase, not on the TV
- Local processing when you don’t want voice audio leaving the house
- Latency under one second from end of phrase to action
- Multi-language support for households that don’t speak English at home
- Background operation that survives screen lock and battery management
- Fallback to text input when voice fails or the room is noisy
Quick comparison
| App | Best for | Platforms | Free plan | Local-only | Rating |
|---|---|---|---|---|---|
| Home Assistant Companion | Direct HA voice integration | Android, iOS | Free | Yes (with local Assist Pipeline) | 4.5 |
| Rhasspy | Self-hosted voice assistant | Android (via HA), Server | Free / Open source | Yes | 4.0 |
| Mycroft AI | Voice satellite hardware | Android (companion), Server | Free / Open source | Yes | 3.8 |
| Willow | ESP32-S3 voice satellites | Web config | Free / Open source | Yes | 4.2 |
| Nabu Casa Cloud | Cloud TTS and STT | Bundled with HA | Subscription | No (cloud) | 4.4 |
| Whisper Voice | Whisper-based local STT | Android | Free / Open source | Yes | 4.0 |
| Wyoming Assist | Wyoming-protocol voice satellites | Server-side | Free / Open source | Yes | 4.2 |
The 7 best Home Assistant voice control apps
1. Home Assistant Companion — best for direct integration
Home Assistant Companion ships the Assist Pipeline as a built-in feature in recent versions. Tap the mic in the app’s header, speak your command, and the phrase is processed by your Home Assistant server’s voice pipeline (local or cloud, your choice). The app also supports background trigger from Assistant on Android in some configurations.
Where it falls short: Background wake-word doesn’t work without a paired voice satellite. The mic button still requires opening the app, which adds friction compared with Alexa’s Echo flow.
Pricing:
- Free / Open source
- Paid: Nabu Casa subscription (modest monthly fee) for hosted TTS and STT
Platforms: Android, iOS
Download: Google Play · App Store · Aptoide
Bottom line: Pick Home Assistant Companion as the starting point for voice control on your phone — every other tool here works alongside it.
2. Rhasspy — best for fully self-hosted voice
Rhasspy is the venerable self-hosted voice assistant framework. It runs as a server (Docker or native) on the Home Assistant host or a separate machine, integrates with HA through MQTT, and supports a wide range of wake words and TTS engines. The Android side is the Companion app pointing at the Rhasspy-driven HA pipeline.
Where it falls short: Setup is more involved than dropping a Nabu Casa cloud config. The newer Wyoming-protocol approach has eclipsed Rhasspy in active development, though Rhasspy still ships updates.
Pricing:
- Free / Open source
- Paid: None
Platforms: Android (via HA Companion), Server (Docker, native)
Download: rhasspy.readthedocs.io (Web-based server)
Bottom line: Pick Rhasspy when you want a tried, fully self-hosted voice stack with MQTT integration.
3. Mycroft AI — best for open voice satellites
Mycroft AI was the open-source voice assistant project that helped popularize self-hosted voice. The Mark II hardware is still around (the company itself folded and was forked by the community as Neon AI), and the Mycroft stack integrates with Home Assistant through dedicated skills.
Where it falls short: The original Mycroft company shutdown left the project fragmented. The Neon AI fork is the most active continuation but moves slower than Wyoming-based options.
Pricing:
- Free / Open source
- Paid: None on the software (Mark II hardware was paid)
Platforms: Android (companion), Server, Mark II hardware
Download: neon.ai (Web-based server, hardware-paired)
Bottom line: Pick Mycroft / Neon AI if you have Mark II hardware or want a long-running open-source voice project.
4. Willow — best for ESP32-S3 voice satellites
Willow is the project that turned the cheap ESP32-S3-BOX hardware into a Home Assistant voice satellite. Wake word, recording, and streaming all happen on the satellite, with inference handled by a local Whisper instance or by Nabu Casa Cloud. Set up costs around $50 in hardware per room.
Where it falls short: Hardware-required (ESP32-S3-BOX or compatible). Setup takes a longer first pass than the official voice preview hardware.
Pricing:
- Free / Open source (software)
- Paid: Hardware cost only (ESP32-S3-BOX)
Platforms: Web config, ESP32-S3 hardware
Download: heywillow.io (Web-flasher, hardware-paired)
Bottom line: Pick Willow when you want cheap-per-room voice satellites and don’t mind the ESP32 setup.
5. Nabu Casa Cloud — best for low-friction TTS and STT
Nabu Casa Cloud is the Home Assistant founder’s commercial cloud service. The subscription delivers high-quality TTS, fast STT, and remote access without port forwarding. Voice satellites can fall back to Nabu Casa when local inference is too slow or unavailable.
Where it falls short: Cloud-dependent for the voice features (the rest of HA stays local). Subscription is a recurring cost.
Pricing:
- Free: No free tier for the cloud (HA core is free)
- Paid: Subscription, modest monthly fee
Platforms: Bundled with HA Companion app
Download: nabucasa.com (Web-based service)
Bottom line: Pick Nabu Casa Cloud when you want the smoothest voice experience without running local inference and you support the HA project at the same time.
6. Whisper Voice — best for local STT on Android
Whisper Voice is one of the standalone Android apps that wrap OpenAI’s Whisper model for offline speech-to-text. Paired with Home Assistant through Tasker or a quick automation, it gives the phone local STT that doesn’t depend on Google’s cloud services.
Where it falls short: Local Whisper inference is slow on older phones. Requires gluing together Tasker or a Webhook to forward transcribed text to Home Assistant — not a turnkey setup.
Pricing:
- Free / Open source
- Paid: None
Platforms: Android
Download: F-Droid (search “Whisper” for the actively maintained build) or sideloadable APK
Bottom line: Pick Whisper Voice when you want local Android STT that doesn’t ship audio to Google.
7. Wyoming Assist — best for protocol-based voice satellites
Wyoming Assist isn’t an Android app but it’s the protocol that Home Assistant’s voice pipeline standardized on. Voice satellites running Wyoming (the official voice preview hardware, the M5Stack AtomS3R, or Willow with a Wyoming bridge) hand audio off to the HA server’s pipeline, which then runs through Whisper for STT and Piper for TTS.
Where it falls short: Configuration happens on the server, not the phone. You’re picking hardware and configuration, not installing an app.
Pricing:
- Free / Open source
- Paid: Hardware cost only
Platforms: Server-side, hardware satellites
Download: home-assistant.io/voice_control (HA docs)
Bottom line: Pick Wyoming Assist as the underlying protocol for any serious local voice setup — most voice satellite hardware now standardizes on it.
How to pick the right one
- For the simplest start: Home Assistant Companion with Nabu Casa Cloud
- For fully local voice without cloud: Home Assistant Companion plus a Wyoming-protocol satellite running local Whisper and Piper
- For multi-room voice satellites on a budget: Willow on ESP32-S3 hardware
- For a tried, self-hosted server stack: Rhasspy
- For local Android STT outside of HA’s pipeline: Whisper Voice
- For long-running open-source voice projects: Mycroft / Neon AI
Most households end up with Home Assistant Companion plus either Nabu Casa Cloud (for ease) or Wyoming satellites with local Whisper (for privacy). Pick a starting point based on which trade-off matters more.
FAQ
Does Home Assistant have voice control?
Yes. Home Assistant ships a voice pipeline (Assist) built into the platform. The pipeline accepts audio from the Companion app, from hardware voice satellites, or from external integrations, runs speech-to-text through Whisper (local) or Nabu Casa Cloud, and triggers Home Assistant intents.
Can Home Assistant voice work offline?
Yes, with the right setup. Whisper STT, Piper TTS, and the Wyoming protocol all run locally on the Home Assistant server. The phone or satellite captures audio, the server processes it, and no voice data leaves the home network.
What’s the difference between Home Assistant Voice and Alexa?
Both let you control smart home devices by voice. Alexa runs in Amazon’s cloud, requires Echo hardware, and ties into Amazon’s broader services. Home Assistant Voice runs on your own server (or Nabu Casa Cloud as an option), uses any compatible microphone or voice satellite, and integrates more deeply with Home Assistant automations.
Do I need a voice satellite for Home Assistant voice control?
Not strictly. The Companion app lets you trigger voice from your phone. But always-on, room-based voice needs dedicated hardware: the official Home Assistant Voice Preview Edition, ESP32-S3-BOX with Willow, or a similar microphone-equipped satellite.
Is Home Assistant voice good enough to replace Alexa?
For controlling Home Assistant entities (lights, switches, scenes, thermostats), yes — and it does so without sending audio to Amazon. For ambient features Alexa is best at (timers, weather, music streaming from Amazon Music), Home Assistant voice still lags. Most households run both for a while during the migration.
