There’s something about the ESP32 that makes people want to create really nice devices with it. Earlier this month, We saw someone build a little desk companion that demands pets.and now, someone else is showing off their latest project: a small puck-shaped device that you can talk to, which responds in kind with text and facial expressions. And it’s so cute.
This ESP32 project is the Tamagotchi that responds
And he has a cute face to boot.
On the ESP32 subreddit, user MRBBLQ posts images and videos of his cute companion. His name is Bopi and he’s designed to clip onto your backpack and wait for you to talk to him, although he may get bored from time to time and decide to show up just to ask you something. The unit is also connected to an expression motor that selects the appropriate emotion for the response.
Here’s how this cool little device works:
Bopi is built on top of LiveKit’s ESP32 SDK, which handles real-time audio streaming over WebRTC directly on the microcontroller. Watcher connects to a LiveKit room, publishes its microphone audio, and subscribes to the agent’s audio; everything runs on the ESP32-S3.
On the server side, a LiveKit agent (agent/) listens to the room. It uses speech to text, an LLM and text to speech to have a conversation. The agent’s responses are transmitted to the Watcher as audio and their transcription is sent as a data stream so that the Watcher can react to individual words in real time.
For facial expressions, the code watches what Bopi says. When it identifies that Bopi says a word that matches one of its expressions, it reproduces the corresponding emotion. It’s a simple little system, but very effective in giving your little one some personality.
Interested in more creator-related content? We launched the weekly XDA Maker newsletter with unique and original content you won’t find anywhere else on XDA. Subscribe by Modifying your newsletter preferences.!
If you want to build your own Bopi (and why wouldn’t you?), you can check out all the necessary hardware and code at the Create GitHub page. There are also an offline version It goes into deep sleep after 30 minutes of no activity and survives for a few days.
