MiniMe — A ChatGPT-Powered Digital Doppelgänger in a Tiny Box, Powered by XIAO ESP32-S3

Michael Merrifield, an emeritus professor of Astronomy in University of Nottingham has pushed the boundaries of AI-powered DIY projects with MiniMe, a pint-sized digital version of himself. Designed to be as compact as possible, this fully self-contained gadget talks like him, thinks like him, and fits neatly in a custom 3D-printed box. And the best part? Much of it was built with help from AI itself — including the code and even the CAD model of the box to put it in!

Hardwares

Here’s what you’ll need to bring your own MiniMe to life:

- Seeed Studio XIAO ESP32S3
- INMP441 I2S microphone
- MAX98357 I2S audio amplifier breakout board
- 3W 8Ω speaker
- 3.7V 300mAh 602030 Li-Po battery or similar (for stand-alone use)
- On-off switch (for stand-alone use)
- Mini breadboard

How It Works

MiniMe combines speech recognition, generative AI, and voice cloning into one seamless, compact device. It starts with voice activation — no buttons necessary. When it hears your voice, the INMP441 microphone captures the audio and sends it to OpenAI’s Whisper API for transcription. That transcribed text is then passed through a customized ChatGPT setup, enhanced with a simple RAG (retrieval-augmented generation) layer. This layer checks a keyword-based database stored on the microcontroller’s SPIFFS filesystem, pulling relevant biographical context and prepending it to the user’s query. The personalized prompt is sent to ChatGPT, which generates a response that’s not just smart, but sounds like “you.” This response is converted back to speech using ElevenLabs’ text-to-speech API — complete with a cloned voice — and finally played aloud through a tiny 3W speaker powered by a MAX98357 audio amplifier.

All of this runs smoothly on the XIAO ESP32S3, which also handles battery management and storage. With no SD card required and everything packed into a tiny 3D-printed enclosure, MiniMe is both a technical feat and a glimpse at the potential of personal AI in your pocket.

To see it in action (MiniMe and Michael talk about astronomy), check out the full demo video below:

Build Your Own MiniMe

Curious to create your own AI twin? Michael has made it easy — the full code, hardware layout, and STL files are available on his GitHub. Whether you’re building a mini-you or a personalized AI companion, the project offers a fascinating glimpse into what AI and microcontrollers can achieve together.

End Note

Hey community, we’re curating a monthly newsletter centering around the beloved Seeed Studio XIAO. If you want to stay up-to-date with:

️ Cool Projects from the Community to get inspiration and tutorials
Product Updates: firmware update, new product spoiler
Wiki Updates: new wikis + wiki contribution
News: events, contests, and other community stuff