⚡ Quick Answer
Home Assistant supports both fully local voice control (via Wyoming protocol + Whisper STT + Piper TTS) and cloud-based voice assistants (Alexa, Google Assistant). The Home Assistant Voice Preview Edition ($59) is the best local voice option — it runs entirely on your LAN with zero cloud dependency. For the easiest setup with the widest device compatibility, use an Amazon Echo Dot ($22–$49) connected via the Alexa integration.
In This Guide
- What's the difference between local and cloud voice control?
- How do you set up fully local voice control in Home Assistant?
- What is the Home Assistant Voice Preview Edition?
- How do you connect Amazon Alexa to Home Assistant?
- How do you connect Google Assistant to Home Assistant?
- Local vs cloud: which is better?
- Our Verdict
- FAQ
What's the difference between local and cloud voice control?
Voice control in Home Assistant falls into two fundamentally different architectures, with very different privacy, reliability, and capability trade-offs.
Cloud voice control routes your voice through a vendor's servers. When you say "Hey Alexa, turn off the living room lights," the audio travels from your Echo to Amazon's servers, is transcribed there, processed, and the command is sent back to your home. This requires an active internet connection and means your voice data is processed by a third party. The upside is near-perfect speech recognition and natural language understanding, plus a huge ecosystem of skills and integrations.
Local voice control processes everything on hardware in your home. With Home Assistant's Wyoming protocol, speech-to-text (STT), natural language processing, and text-to-speech (TTS) all run on your LAN. No audio ever leaves your home. The trade-off is that local AI models are less capable than cloud models, and you need capable hardware to run them with acceptable latency (typically under 2 seconds with a modern CPU or GPU).
How do you set up fully local voice control in Home Assistant?
Home Assistant's local voice pipeline uses three components: a wake word engine, a speech-to-text model, and a text-to-speech engine. Here's how to set it up:
- Install the Wyoming Protocol add-on: In HA, go to Settings → Add-ons → Add-on Store. Install "Wyoming Satellite" if you're using a satellite device, or the individual Whisper and Piper add-ons for server-side processing.
- Install Whisper (STT): The Whisper add-on runs OpenAI's Whisper model locally. The "tiny" model processes speech in ~0.5 seconds on a Raspberry Pi 5; the "base" model is more accurate but takes ~1.5 seconds.
- Install Piper (TTS): Piper generates natural-sounding speech locally. Choose a voice pack (English US, English GB, and dozens of other languages are available).
- Set up a wake word: The "openWakeWord" add-on listens for phrases like "Hey Jarvis," "OK Nabu," or "Hey Mycroft" entirely locally.
- Configure the Voice Pipeline: Go to Settings → Voice Assistants → Add Assistant. Select your Whisper STT, Piper TTS, and wake word engine. This creates a named voice pipeline you can assign to satellite devices.
Hardware requirements: A Raspberry Pi 4 (4GB) can run Whisper tiny/base adequately. A Raspberry Pi 5 (8GB) or a machine with a GPU runs the "small" Whisper model smoothly. An Intel NUC or mini PC runs the "medium" model in real time.
What is the Home Assistant Voice Preview Edition?
The Home Assistant Voice Preview Edition ($59) is an ESP32-S3-based satellite device built by Nabu Casa. It connects to Home Assistant via the Wyoming Satellite protocol over Wi-Fi and passes audio to your HA instance for processing. It features:
- Dual microphone array for better far-field pickup (effective range ~3–4 metres)
- Speaker output for TTS responses
- LED ring for visual status feedback
- 100% open source firmware (ESPHome-based)
- Fully local operation — no cloud, no subscription
Setup takes about 15 minutes: flash the ESPHome firmware, connect to Wi-Fi, and it auto-discovers in Home Assistant as a Wyoming satellite. Assign your local voice pipeline and it's ready. At $59, it's more expensive than an Echo Dot but every word you say stays on your LAN.
Home Assistant Voice Preview Edition — Best Local Voice Hardware
Purpose-built for local, private voice control with Home Assistant. No cloud, no subscription, fully open source. Ships with ESPHome pre-flashed and auto-discovers in HA within minutes.
Check Price on AmazonHow do you connect Amazon Alexa to Home Assistant?
The easiest way to connect Alexa is via Home Assistant Cloud (Nabu Casa) at $6.50/month. This handles all the AWS Lambda setup automatically. Alternatively, you can configure it manually for free using the Alexa Media Player HACS integration and your own AWS account — but this takes 1–2 hours and requires AWS configuration.
Via Nabu Casa (recommended):
- Subscribe to Home Assistant Cloud at nabu.casa ($6.50/month)
- Enable the Alexa integration in HA Cloud settings
- In the Alexa app, enable the "Home Assistant" skill and link your Nabu Casa account
- Say "Alexa, discover devices" — your HA entities appear in Alexa within 60 seconds
You can control which entities Alexa can see by filtering them in the Alexa integration settings in Home Assistant. Supports lights, switches, sensors, climate, covers, locks, and media players.
Amazon Echo Dot (5th Gen) — Best Cloud Voice Hardware for HA
The Echo Dot 5th gen ($49.99, often on sale for $22–$35) offers excellent microphone pickup, Alexa integration, and a compact design. It's the best value entry point for cloud-based voice control with Home Assistant.
Check Price on AmazonHow do you connect Google Assistant to Home Assistant?
Google Assistant integration also works best via Home Assistant Cloud. Enable it in Nabu Casa settings, then in the Google Home app link the "Home Assistant" action. Google Home can control lights, switches, thermostats, locks, and media players exposed by HA. Voice commands like "Hey Google, set the bedroom to 20 degrees" work natively.
For free/manual setup, the Google Assistant SDK and Home Graph integrations require a Google Developer account and more complex configuration — but it's free beyond the HA subscription.
Local vs cloud: which is better?
| Feature | Local (Whisper/Piper) | Cloud (Alexa/Google) |
|---|---|---|
| Privacy | 100% local, no data sent out | Audio processed on vendor servers |
| Works without internet | Yes | No |
| Recognition accuracy | Good (Whisper small/medium) | Excellent |
| Setup complexity | Moderate (30–60 min) | Easy (15 min with Nabu Casa) |
| Cost | Free (hardware cost only) | $6.50/month (Nabu Casa) or free manual |
| Hardware needed | Capable HA host + satellite | Just an Echo/Nest speaker |
| Custom commands | Full HA template support | Limited to supported entity types |
| Response speed | 0.5–2 seconds | ~0.3–0.8 seconds |
Our Verdict
For privacy-conscious users who want zero cloud dependency, the Home Assistant Voice Preview Edition ($59) running a local Whisper pipeline is the best choice. For everyone else who wants the easiest setup and best accuracy, an Echo Dot ($22–$49) connected via Nabu Casa ($6.50/month) is the pragmatic winner. The ideal setup for many homes is both: local voice satellites in bedrooms, Alexa/Google in shared spaces.
Frequently Asked Questions
Can I use voice control without a subscription?
Yes — the fully local voice pipeline (Whisper + Piper + openWakeWord) is completely free. For cloud integrations like Alexa and Google Assistant, the manual setup method is also free but requires technical configuration. Only the Nabu Casa shortcut costs $6.50/month.
What languages does local voice control support?
Whisper supports over 90 languages including Spanish, French, German, Greek, Japanese, and many others. Piper TTS has voice packs for 30+ languages. English recognition is most accurate; other languages depend on available model quality.
Can I use my existing Echo or Google Nest speakers?
Yes. If you have Alexa or Google Assistant devices, connecting them to Home Assistant via Nabu Casa (or the manual method) lets you use them as HA voice controllers. You don't need to buy new hardware.
How accurate is Whisper for voice control?
Whisper's "small" model achieves approximately 92–95% word-error-rate accuracy on clear English speech, which is sufficient for most home automation commands. The "medium" and "large" models approach cloud-level accuracy but require significantly more compute.
Can I run voice control on a Raspberry Pi?
Yes. A Raspberry Pi 4 (4GB) runs Whisper tiny/base with 1–2 second latency. A Raspberry Pi 5 (8GB) handles the base/small model more comfortably. For the medium model, an x86 machine is recommended.
Does local voice control work during internet outages?
Yes, that's the primary advantage. Local voice control works as long as your LAN and Home Assistant are running, regardless of internet connectivity. This is especially valuable for homes that experience frequent outages.
SmartWired participates in the Amazon Associates Programme. We may earn a commission from qualifying purchases at no extra cost to you.