A open-source voice AI assistant

Hardware：https://oshwhub.com/wlmh110/esp32-i2s-tf

Software：https://github.com/wlmh110/ESP32test

The hardware cost is approximately $6 (excluding battery and processing fees). Key components include an ESP-32 ($2.5) and an INMP1441 ($1.5).

Hardware verification for the proposed solution has been completed. The ESP32-based hardware used for verification has been validated. This verification board can record audio upon button press and stop recording upon release. The recorded data is then sent to a TCP server.

This verification board can now serve as a basic trigger-based voice assistant, although it is still far from our ultimate goal. Attached are photos of the verified hardware solution. The open-source link for the entire project design will be updated shortly, including hardware design, preliminary verification code, and a simple test server code (which merely packages the received audio data into a WAV file).

SMT is completed, and next we will test the function of the recording part.

The design uses the ESP32-WROVER primarily because of its large memory capacity. 

The MIC is an INMP441, and data is directly obtained via I2S. 

There is a button for user operation and an RGB LED to indicate status. 

There is a TF card slot, which is not part of the original functionality because the TF card consumes a lot of power; it is included only for debugging audio data. 

The battery uses a PH2.0 connector to facilitate the use of different solutions. 

The PCB has a diameter of 4 cm and includes a slot for the ESP32 antenna.

The current hardware cost of the solution is approximately $6 (excluding the battery).

Since the PCB has not been verified yet, a public link for the schematic diagram generation has not been created. After the verification is completed, the schematic diagram and PCB address will be filled in the project.