Custom Wake Word Request for ESP32-S3-BOX-3: "OK Medi"

[Bndsz99]

Checklist

• Checked the issue tracker for similar issues to ensure this is not a duplicate.
• Provided a clear description of your suggestion.
• Included any relevant context or examples.

Issue or Suggestion Description

Hello Espressif Team,

I am developing a healthcare assistant project called "Medi Assistant" using the ESP32-S3-BOX-3 hardware. To make the device more intuitive for users, I would like to implement a custom wake word: "OK Medi" (or alternatively "Hey Medi").

I am currently using the esp-sr framework, but as a developer, I find the process of creating a high-quality, custom wake word model (WN binary) somewhat restrictive for specific non-standard phrases in regional contexts.

Technical Details:

Hardware: ESP32-S3-BOX-3

Framework: ESP-IDF v5.5.4

Component: esp-sr / WakeNet

Primary Language: English (with potential Hungarian phonetic influence)

Questions / Requests:

Model Generation: Is it possible to get assistance in generating a .bin model for "OK Medi" compatible with WakeNet 8 or 9?

Training Pipeline: Are there any updates regarding a localized or offline tool for developers to train simple custom wake words without the full Voice Customization service process?

Optimization: Given the dual-mic setup of the BOX-3, are there specific AFE (Audio Front-End) configurations you recommend to maximize the detection accuracy for this specific phrase?

I believe that enabling more accessible custom wake words would greatly benefit the developer community using the ESP32-S3-SENSE and BOX series.

Thank you for your incredible work on the ESP-SR framework!
Andras
bandesz@iboss.hu