I came up with the idea of hosting an AI model on a local server in my house in order to decrease dependency on Google and other services. A dedicated device is used in the setup to captures spoken speech, convert it to text using voice recognition software, and then transfer the text to an AI model for processing. After analyzing the text, The AI model would analyze the text and generate responses, which are then converted back into speech and played aloud. This system allows for a private, self-contained AI interaction environment without depending on external cloud services that i don't trust.
I began by dismantling a low-quality, free speaker to extract its speaker components. Once I had the measurements, I designed a simple enclosure using Fusion360. This basic design was intended for initial 3D printing tests to ensure proper fit and functionality before proceeding with more advanced iterations
Once I had the correct measurements and a rough design, I ordered the necessary components from AliExpress.
After receiving all the components and assembling the system, the initial startup attempt was unsuccessful. Unfortunately, due to incorrect wiring, the ESP32 microcontroller was damaged and rendered unusable. The mishap highlighted the need for careful attention to wiring and connections during the assembly process.
Frustrated by the initial failure, I took a break from the project. In the meantime, I refined the enclosure by placing a sock over the speaker to conceal it and enhance the final design's appearance. This adjustment aimed to improve the aesthetic of the enclosure while I awaited renewed interest in the project.