Stack-chan minimal - voice in English

To make Stack-chan popular worldwide, I am creating a gachapon-sized gadget called the "AI Stack-chan minimal series." This product doesn't have AI features yet, but as a step towards making it multilingual, I've developed a version that speaks English with the adorable Stack-chan face (M5Avatar). In near future, this product is going to add AI chat bot functions based on Local LLM.

Technical Introduction

1. arduino-espeak-ng as TTS Solution

First, I would like to express my gratitude to pschatzmann for publishing the arduino-espeak-ng library.

The M5Stack Atom series is based on the ESP32 chip, but there are limited TTS libraries available for ESP32 that can deliver satisfactory audio quality. Initially, I created a version that could run on SRAM only, but the audio quality didn't meet my family's expectations.

Since arduino-espeak-ng comes with multiple expressive character voice data, I switched to the AtomS3R and modified the implementation to utilize both PSRAM and PROGMEM for character voice playback. The source code is publicly available, so feel free to experiment with different characters and languages!

Resources:

• pschatzmann/arduino-espeak-ng https://www.pschatzmann.ch/home/2022/11/18/espeak-ng-setting-the-voice/
• Character Data: https://github.com/pschatzmann/arduino-espeak-ng/tree/main/espeak-ng-data/voices/!v

2. Lip Sync Integration with M5Avatar

To ensure good usability for other projects, I attempted to implement the simplest possible code using M5Unified. However, the M5Speaker class had compatibility issues with arduino-espeak-ng, and both Claude and Gemini AI told me that achieving lip sync synchronized with audio was impossible.

After reviewing the source code extensively, I finally got it working properly. While the voice quality is excellent, watching the Stack-chan Minimal speak with lip sync is truly heartwarming!

3. Next Steps

Currently, I'm developing a Japanese version of AI Stack-chan Minimal. For the Maker Faire Tokyo 2025 exhibition in October, I plan to showcase an AI Stack-chan Minimal that can speak in both Japanese and English.

Upcoming Events:

• Maker Faire Tokyo 2025: https://makezine.jp/event/mft2025/
• AI Stack-chan Minimal: https://x.com/UtaAoya/status/1947872269361156375