Development History | AI Companion Toy Development (1)
Developing AI companion toys, revealing the whole process from replicating open source projects to fully self-developed hardware.
Core content:
1. Replicating the "ESP32 Xiaozhi" open source hardware project to accumulate AI technology
2. Demonstration of the whole process of hardware assembly, firmware burning, network configuration, and binding
3. Power solution optimization, the transition from 18650 lithium battery to polymer lithium battery
This year, we want to accumulate some technologies in the field of AI. We will first replicate the open source hardware project "ESP32 Xiaozhi" to start with, and then gradually build an AI companion toy hardware product solution with fully self-developed hardware, firmware and cloud.
Replicate Xiaozhi's interim results: Development board testing completed.
The hardware includes: ESP32s3 development board, INMP441 microphone, MAX98357A amplifier, 4 ohm 3W speaker, 0.91 inch OLED, lithium battery, charging module, volume button, function button, power switch.
Firmware: Directly download the firmware of the Xiaozhi development board and burn it.
Cloud AI: You need to log in to xiaozhi.me, register an account, and manage devices. For the first device access, you need to add a new device and enter the verification code broadcast by the device. After that, you can do some general configuration and maintenance on the device's AI agent in the device management interface.
Xiaozhi demonstration video:
I went through the process of hardware assembly, firmware burning, network configuration, and binding. I have to say that Xia Ge's Xiaozhi solution is definitely the best AI hardware entry solution on the entire network. The teaching material demonstration steps are very smooth, the copywriting is clear and concise, and the pictures are simple and easy to understand. Developers with embedded experience will be able to run through the entire process within 1 hour if the information is complete, and then start their own creation. As long as a novice without embedded experience has some hands-on ability, he should be able to successfully assemble a personal AI assistant.
In order to make the product later, I added a charging module to the original solution. Initially, I wanted to use the 18650 lithium battery + TP4056 solution, but later I found that this power supply solution had problems. 3.3V could not drive the entire circuit and needed to be boosted to 5V for power supply. In addition, 18650 was large and bulky and not suitable for AI companion products that can be played with, so I changed to a power supply solution using a polymer lithium battery + IP5306. After verification, there was no problem with the power supply and charging of the power module, and the development board could also operate normally.
