Yes—if every part of the voice-control chain has a local path. Home Assistant can process speech on your own hardware and control devices over a working home network while the internet connection is down. That does not make cloud-only lights local: a bulb, appliance, speech engine, remote-access service, or integration that requires a vendor server will stop working when that server is unreachable.
What “offline” means in a smart-speaker setup
An internet outage and a local-network outage are different events. Your router, Wi-Fi access point, Ethernet switches and Home Assistant host can continue operating on the local network even when the WAN connection is disconnected. In that situation, a local voice endpoint can still send audio to Home Assistant, Home Assistant can interpret the command, and a locally connected device can respond.
As an Amazon Associate I earn from qualifying purchases.
If the local network itself is down, the result is different. The microphone cannot reach the server, the server cannot reach the device, and a wireless satellite cannot return a spoken response. Offline voice control therefore means “no internet dependency,” not “no network of any kind.”
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Home Assistant’s documentation says its own server can keep running without internet and identifies Zigbee, Z-Wave, Matter, Thread and ESPHome as examples of local device protocols. The exact integration still matters: not every integration offered by Home Assistant is local. See Home Assistant’s offline FAQ for the documented boundary.
#1 Best Overall
- Designed for Home Assistant Voice & Music Workflows: Preloaded with Home Assistant Voice Assistant and Music Assistant. Functions as both a voice input terminal and an audio playback endpoint.
- Dual Microphones for Voice Capture: Built with dual digital microphones for wake word or button-activated voice capture. Audio is streamed to the Home Assistant voice pipeline.
- Integrated 3W Speaker for Direct Playback: The built-in 3W/4Ω speaker supports TTS playback, Music Assistant streaming, and system audio without external speakers.
- Linux-Based Local Operation: Runs a lightweight Linux system on a quad-core ARM A53 CPU with 256MB RAM and 512MB flash for local audio processing.
- Development & Debugging Capabilities: Supports firmware flashing, and also provides access to live logs, on-device editing—suitable for routine development or issue diagnosis.
The local voice-control chain
A command such as “turn off the hallway light” is not performed by a speaker alone. It passes through several components, and each must be available locally for the command to remain offline.
- Endpoint: A phone, dedicated satellite or DIY microphone/speaker device captures the voice and plays the response.
- Speech-to-text: A local recognizer converts the audio into words.
- Intent handling: Home Assistant maps those words to an action, such as turning off a specific light.
- Device integration: Home Assistant sends the command through a local protocol or local API.
- Text-to-speech: A local voice engine can speak a confirmation or answer.
Home Assistant calls this its Assist pipeline. You must expose an entity to Assist before it can be controlled by voice; an entity that is hidden from the voice assistant is not available to commands. The fully local voice guide explains the pipeline and supported listening endpoints.
This chain also explains why “the speaker works offline” is an incomplete statement. A local microphone and local speech recognition cannot control a cloud-only appliance. Conversely, a locally integrated light can remain controllable even if a particular voice endpoint is replaced with another one.
Which devices can be controlled without internet?
Devices with local protocols
Choose products whose Home Assistant integration documents local communication or whose protocol is designed for local control. Zigbee, Z-Wave, Matter, Thread and ESPHome are common examples, although the controller hardware and integration configuration determine whether the installation is actually local. A Matter device may still depend on a local Matter controller; a Thread device requires a functioning Thread border-router arrangement.
Before buying or migrating a device, check the exact model and integration—not just the marketing name of the product family. Confirm that basic commands such as on/off, brightness, temperature or lock state are available locally and identify what happens after a power cycle or controller restart.
Cloud-only devices
A device that accepts commands only through its manufacturer’s internet service cannot be controlled when that service is unavailable. It may appear in Home Assistant, but the integration is then a cloud bridge rather than a local path. The same limitation applies if the vendor requires an online account for every command.
Remote and assistant integrations
Remote access through Home Assistant Cloud, and integrations with Apple Home, Google Home or Amazon Alexa, are internet-dependent features. They can be useful when you are away from home, but they are not part of an internet-free control path. Automatic updates, app-store downloads and some third-party services also stop during an outage.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Local speech recognition: speed versus flexibility
Home Assistant documents two important local speech-to-text choices. The right one depends on the commands you expect and the computer running Home Assistant.
| Recognizer | Best fit | Published performance examples | Trade-off |
|---|---|---|---|
| Speech-to-Phrase | Frequent, predictable home-control commands | Under one second on Home Assistant Green or a Raspberry Pi 4 | Closed-ended command set; it is not intended for unrestricted conversation |
| Whisper | More open-ended speech and capabilities beyond simple control | Around eight seconds on a Raspberry Pi 4; under one second on an Intel NUC | Needs more capable hardware for responsive results |
Those times are the figures Home Assistant reports for the named hardware examples, not a universal latency guarantee. Microphone quality, language model, recording length and concurrent workloads can change the result. Test on the host you actually plan to keep running.
For local responses, Piper provides text-to-speech. Home Assistant reports that Piper can generate 1.6 seconds of speech in one second on a Raspberry Pi when using medium-quality models. Voice quality and language availability vary by language, so verify the language and voice you need before designing the system around it. The detailed setup is in Home Assistant’s fully local voice-assistant guide.
Hardware roles: server, endpoint and device controller
The always-on Home Assistant host
Home Assistant needs a machine that remains powered. The official hardware guidance lists Home Assistant Green, a Raspberry Pi, a mini PC or NUC, a virtual machine and other supported systems. Green is a plug-and-play host option; a Raspberry Pi or mini PC may suit an existing setup, while a virtual machine requires a suitable computer and virtualization configuration. See Home Assistant hardware options and the hardware FAQ.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallThe host is the automation brain, not automatically a room microphone. You still need an endpoint in the room where you want to speak.
Rank #3
Phone or tablet endpoint
The Home Assistant companion app can act as a listening endpoint. This avoids buying another microphone, but the phone must be on the local network, have microphone permission, and remain available when you want to issue a command.
Dedicated voice satellite
The Home Assistant Voice Preview Edition is a room endpoint rather than a standalone Home Assistant server. Its product page lists a dual-microphone array, speaker, USB-C power, Wi-Fi, Bluetooth, an audio output, ESPHome firmware, and a physical mute switch that cuts microphone power. Its ESP32 and XMOS firmware is open source. It still requires Home Assistant running on another device and a configured Assist pipeline. Details are on the official Voice Preview Edition page.
The product documentation distinguishes focused local processing of common home-control phrases from full local processing. More capable local processing needs substantially stronger host hardware to maintain useful speed and accuracy. Do not treat the endpoint as a promise of unrestricted, fast offline conversation.
Recommended Free Tools
DIY ESPHome satellite
ESPHome-compatible voice devices can stream microphone audio to Home Assistant for processing and play returned audio through a speaker or media player. Wake-word support can be included, but voice and audio components consume significant RAM and CPU on a microcontroller. ESPHome warns that resource-heavy combinations can crash, especially when other components are enabled. Voice Assistant support requires Home Assistant 2023.5 or later according to the current ESPHome Voice Assistant documentation; treat that as a compatibility threshold, not a recommendation to run an old version.
Linux and other experimental endpoints
Home Assistant’s Assist overview describes an experimental Linux Voice Assistant route for compatible x64 or ARM64 hardware. It can be useful when you have spare hardware and are comfortable troubleshooting audio devices, services and permissions. Experimental status means compatibility and setup details can change.
Wake words and privacy controls
A wake word can keep the microphone from sending every conversation to the assistant pipeline, but support depends on the endpoint and current software. The official wake-word guide says Voice Preview Edition does not currently support custom wake words. If a custom phrase is essential, check the current device-specific guidance at Home Assistant’s wake-word documentation before choosing hardware.
Rank #4
- Fitment For: Designed for Home Assistant Voice Preview Edition, ensuring a secure and stable wall mounting fit.
- Optimized Voice Interaction: Positions your device at a convenient height for better voice control experience.
- Stable Device Support: Provides firm support to keep the device securely positioned during daily use.
- Space Saving Wall Mount: Wall mount holder keeps your smart speaker off desks and counters for a cleaner setup.
- Multi Room Application: Ideal for kitchens, bedrooms, offices, and other smart home locations.
Local processing narrows the audio privacy boundary: speech can stay inside your home network instead of being uploaded to a remote speech service. It does not eliminate every privacy consideration. Review which integrations are enabled, whether remote access is turned on, what logs are retained, and whether any optional cloud feature is active. A physical microphone mute that removes power, as listed for Voice Preview Edition, gives a clear hardware-level control.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Clear out junk files and repair common Windows errors3Scan for outdated or missing drivers - takes under a minuteHow to build a fully local setup
- Keep the local network available. Use a reliable router and Wi-Fi or Ethernet coverage. An internet outage is acceptable; a failed LAN is not.
- Install Home Assistant on an always-on host. Select Green, a Raspberry Pi, mini PC/NUC, virtual machine or another supported platform. Follow the installation method for that platform.
- Add and test device integrations. Pair local Zigbee, Z-Wave, Matter, Thread or ESPHome devices and operate them from the Home Assistant interface while disconnected from the internet. If a device stops responding, determine whether its integration is cloud-dependent.
- Install local speech-to-text. Use Speech-to-Phrase for a constrained set of fast home commands, or Whisper when you need broader language handling and have sufficient compute.
- Add local text-to-speech. Configure Piper and select a supported voice and language. Check the resulting audio from the actual host rather than assuming every voice has identical quality.
- Choose an endpoint. Use the companion app, a dedicated satellite or an ESPHome device. Confirm the endpoint can reach the Home Assistant host by its local address.
- Expose entities to Assist. In Home Assistant’s voice-assistant settings, expose only the lights, switches, scenes and other entities you want voice control to access. Give them clear room and device names.
- Configure the wake word if needed. Verify that the selected endpoint supports the wake-word option you want; do not assume custom wake words are available.
- Test with the internet disconnected. Unplug the router’s WAN connection while leaving the local network powered. Test a simple command, a multi-room command, a response from Piper and each device category you depend on.
- Document recovery. Record the Home Assistant host address, endpoint locations, device pairing details and the last known-good software versions so you can restore the system without an online service.
Open-source alternatives beyond Home Assistant
Rhasspy describes itself as an open-source, fully offline collection of voice-assistant services that works with Home Assistant and other home-automation platforms. Its architecture can split processing between a more powerful base station and remote audio satellites, which is useful when room endpoints have limited computing resources. The project describes technically capable users as its audience. Its documentation does not establish a current release cadence or a support guarantee, so evaluate maintenance and compatibility before making it the foundation of a new installation. Start with the Rhasspy project page and “Why Rhasspy?” documentation.
Choosing a build path
| Path | Processing location | Hardware roles | Command range and speed | Setup and upkeep | Best for |
|---|---|---|---|---|---|
| Phone plus local Home Assistant | Local speech and automation on your host | Existing host plus phone | Depends on Speech-to-Phrase, Whisper and host performance | Least additional hardware; phone permissions and availability matter | Trying local voice control with equipment you already own |
| Dedicated local satellite | Endpoint captures and plays audio; host performs the pipeline | Always-on host plus room satellite | Predictable room access; processing capability depends on host | More convenient than DIY, but endpoint still needs configuration and power | Several rooms or a permanent microphone and speaker |
| ESPHome DIY satellite | Audio streams to local Home Assistant | Host plus compatible ESP32 hardware, microphone and speaker | Suitable for supported commands; microcontroller resources constrain features | Highest configuration and troubleshooting burden | Technically inclined users who want customization |
| Rhasspy-style distributed setup | Offline services can run on a base station with remote satellites | Powerful base station plus audio endpoints | Depends on configured services and hardware | Flexible, but project status and compatibility require checking | Users comfortable assembling an offline voice stack |
Use five checks before committing: processing location, verified device connectivity, hardware roles, command range and speed on your host, and setup plus privacy controls. A cheap endpoint paired with a cloud-only device is not an offline system; a more capable server paired with local devices can be.
Common failure modes and fixes
“The command is heard, but nothing happens”
- Confirm the target entity is exposed to Assist.
- Check the entity’s integration and whether it requires a vendor cloud.
- Test the same device from the Home Assistant dashboard while the internet is disconnected.
“The assistant is too slow”
- If you are using Whisper on a Raspberry Pi 4, Home Assistant’s published example is around eight seconds; move processing to a stronger host such as an Intel NUC-class machine or use Speech-to-Phrase for constrained commands.
- Reduce unnecessary audio components on ESPHome hardware and review RAM and CPU use.
- Check whether long recordings or an unsuitable language model are increasing processing time.
“The endpoint cannot connect”
- Verify that the endpoint and Home Assistant host are on the same reachable LAN or VLAN.
- Check Wi-Fi power, local DNS or host address, and firewall rules.
- Remember that a working internet connection cannot repair a broken local route, and an internet outage does not by itself imply a local-route failure.
“It worked until the internet was unplugged”
- Identify whether speech recognition, text-to-speech or the device integration was still calling a cloud service.
- Disable or replace that component with a local alternative.
- Retest after a clean restart, because some integrations cache credentials or state only temporarily.
What still needs internet access
A local installation is not the same as an internet-free setup process forever. Initial downloads, operating-system and Home Assistant updates, app-store access, remote access, cloud integrations and some language or AI features may require internet connectivity. Plan updates and backups when the connection is available, and do not assume that a locally functioning command path provides remote control while you are away from home.
Bottom line for a buying or build decision
Open-source smart speakers can control devices without the internet when the entire path is local: an always-on Home Assistant host, local speech-to-text and text-to-speech, a reachable endpoint, and devices controlled through verified local integrations. Speech-to-Phrase is the sensible low-latency choice for familiar commands; Whisper offers broader recognition but needs more compute. A dedicated satellite or DIY ESPHome endpoint supplies the microphone and speaker, while Home Assistant remains the server. Verify each device model, language, wake-word feature and integration before purchase, then test with the WAN link disconnected but the home network running.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
