Willow

Willow is an open source, self-hosted voice assistant platform that runs on ESP32-S3-BOX hardware to provide private, high-speed command recognition.

Willow is an ESP IDF-based, privacy-focused voice assistant platform designed for the ESP32-S3-BOX hardware family. It provides local speech recognition and command processing, allowing users to keep all data on their own networks without internet dependency. The system functions as an alternative to commercial voice assistants, offering response times of 500ms or less from the end of speech to completed action. It integrates with Home Assistant, openHAB, and generic REST APIs to execute commands based on voice input. Users can either use local command recognition for up to 400 commands or connect to the open source Willow Inference Server for full speech transcription. The hardware setup is self-hosted and requires approximately $50 of commercially available components.

Willow at a glance

Best for: Software engineer
Runs on: self-hosted
Pricing: Approximately $50 hardware cost.
Released: March 2023
Category: AI audio tool
Website: heywillow.org

Key features

Local command recognition for up to 400 commands
Integration with Home Assistant, openHAB, and REST APIs
Support for multiple wake words
Automatic gain control and noise suppression
Far-field voice recognition performance
Audio compression for 2.4 GHz Wi-Fi environments
Touchscreen and LCD user interface support

Use it for

Automating home functions via Home Assistant
Voice-controlling openHAB-connected devices
Transcribing speech via Willow Inference Server
Executing REST API calls using voice commands

Alternatives to Willow

AssemblyAI — Industry-leading speech models deliver high accuracy and low latency for voice applications.
Piper — Runs locally and offers fast neural text-to-speech synthesis.
Speaches — It acts as an Ollama equivalent for running local TTS and STT models.
Esp32 Thinks — Runs your own AI stack on low-cost hardware without vendor lock-in.
Pipecat — Builds real-time voice and multimodal agents with frame-level control.
sherpa-onnx — Runs complex audio AI tasks offline on local devices and embedded hardware.
Vosk — It enables offline, low-latency speech recognition on small devices.
Esp32 Pie ASR — Faster than real time streaming speech recognition on a microcontroller alone
WeNet — Achieves state-of-the-art results on public speech datasets.
Qwen Audio Agent — Separates voice conversation from task execution for continuous interaction.
Free ASR — Zero-dependency real-time streaming speech-to-text with no account login required.
whisper.cpp — It provides high-performance, offline speech recognition on diverse hardware.

All Willow alternatives