Show HN: Willow – Open-source privacy-focused voice assistant hardware (github.com)

Article summary

The article introduces Willow, an open-source, privacy-focused voice assistant hardware that can be self-hosted. Willow uses the ESP BOX and ESP SR speech recognition framework for wake word detection and speech recognition. The device supports up to 400 commands and can be integrated with Home Assistant for tighter integration and dynamic configuration. The creators plan to release an open-source inference server next week.

Main themes

• voice assistant
• privacy
• security
• speech recognition
• home automation
• open-source hardware
• inference server
• microphone array
• noise cancellation
• device control

What commenters say

• Some users are interested in the project due to privacy concerns with existing voice assistants.
• The device's speech recognition accuracy and response latency are comparable to commercial products like Alexa.
• There are potential security risks with the device, but the creators have taken steps to mitigate them, such as using WPA3 and protected management frames.
• The use of Whisper for speech recognition is sufficient for the application, despite its limitations in streaming ASR.
• Some users are interested in using the device with other models, such as Vicuna 13 or LLaMA.
• The device's microphone array and noise cancellation capabilities are effective in noisy environments.
• There are discussions about the potential for the device to be hijacked and used for anti-privacy purposes, but the creators believe it is secure.
• Some users are interested in the project's potential for controlling devices and performing tasks, and believe it could cover a significant portion of their current voice assistant use.