esp-srmodels

Pre-built ESP-SR speech recognition model partition images for Willow devices.

What is this?

This repository provides ready-to-flash srmodels.bin files for Willow's OTA model update feature. Each release contains individual wake word model partition images built from Espressif's ESP-SR repository.

Downloads

Check the Releases page for pre-built model tarballs.

Each tarball contains:

• srmodels.bin — packed model partition image (flash via Willow OTA)
• LICENSE — MIT license from ESP-SR
• SOURCE.txt — link to the original model source

How it works

A GitHub Actions workflow:

1. Clones the ESP-SR repository
2. Packs each WakeNet v9 (wn9_*) model using Espressif's pack_model.py
3. Creates tarballs with license attribution
4. Publishes as a GitHub Release

License

The wake word models are from Espressif's ESP-SR and are licensed under the MIT License.