Home Assistant Voice Preview Edition

The Home Assistant Voice PE is a local voice assistant for Home Assistant that focuses on privacy, openness, and local processing. It is not an Alexa or Google replacement in the sense of "plug-and-play with cloud AI," but rather a system intentionally designed for local operation, whose performance depends heavily on the Home Assistant hardware being used.

Classification & Expectation Management

Voice quality, recognition accuracy, and response time depend significantly on CPU power, memory, and the selected language model. Commercial assistants like Alexa or Google Assistant rely on scaled-up cloud infrastructure with specialized AI accelerators. In contrast, the Home Assistant Voice PE operates locally or in a hybrid mode—with correspondingly more realistic expectations regarding latency and speech recognition.

Features

• Local voice control via Home Assistant Assist.
• ESPHome pre-installed; fully open firmware (ESP32-S3 and XMOS).
• Optimized audio processing: echo cancellation, noise reduction, auto-gain control.
• Hardware Mute: physical switch disconnects power to the microphone.
• Grove port and exposed pads on the circuit board for extensions and modding.
• Internal speaker and dual-mic array.
• Compact plastic housing (white / semi-transparent), screwed together and easy to open.

Technical Specifications

• SoCs: ESP32-S3 with 16 MB flash and 8 MB PSRAM.
• Audio processing: XMOS XU316.
• Audio output: 3.5 mm jack, TI AIC3204 DAC with 48 kHz sampling rate.
• Network: 2.4 GHz Wi-Fi and Bluetooth 5.0 LE (no Ethernet).
• Power supply: USB-C, 5 V, 2 A.
• Dimensions: 84 × 84 × 21 mm; weight: 96 g.
• Material: Polycarbonate plastic housing, white and semi-transparent.
• Controls: Multi-function button, rotary knob, hardware mute switch.

Requirements & Hardware Recommendations

• A pre-configured Home Assistant system on the same network.
• For fully local voice processing (STT + intent recognition) recommended:
  • Intel N100 / N200 Mini-PC
  • Intel i3 (8th Gen or newer)
  • AMD Ryzen 3 / 5 (Zen 2 or newer)
  • At least 8 GB RAM, 16 GB preferred
• Raspberry Pi 4/5: limited suitability; local models react more slowly and with reduced recognition quality.
• For weaker hardware: significantly better results with Home Assistant Cloud (hybrid approach).

Voice Quality – A Realistic Assessment

• Good recognition with clear pronunciation and a quiet environment.
• Noticeably more sensitive to background noise than Alexa devices.
• Complex or very open-ended voice commands require cleanly configured intents.
• Strengths lie clearly in defined smart home commands, not in free conversation.

Installation

1. Power the device via USB-C.
2. Connect to the local Wi-Fi (2.4 GHz).
3. Automatic detection in Home Assistant via the ESPHome integration.
4. Set up Assist and select language models suitable for the available hardware.

Special Features

• Full privacy control via physical microphone power disconnect switch.
• Open system with no forced manufacturer cloud.
• Grove port for additional sensors or extensions.
• Suitable for single rooms or distributed Assist setups with multiple devices.