This needs to be a Raspberry Pi 4, so that you can run the 64-bit operating system. Earlier versions of Raspberry Pis are likely to throw memory errors while running this program. The 2 GB RAM model is sufficient.
×
1
Raspberry Pi power supply
I recommend the official Raspberry Pi 5v power supply.
×
1
USB Microphone
For talking to the assistant. Any USB mic should work.
×
1
USB Speaker
For the assistant to talk back to you.
×
1
LEDs (optional)
The program optionally lights LEDs when it detects the wake word and pulses them when it is responding to your queries. Spherical LEDs were used for DaVinci's eyes in the video demo.
×
1
Resistors (220 ohm) (optional)
Only needed if you decide to use the optional LEDs.
×
1
Heatsinks (optional)
The heatsinks are optional but are highly recommended if you will put your Raspberry Pi in a case or other confined container such as the one used in the video demo.
×
1
USB Splitter Cable (optional)
If you decide to print and use the DaVinci body, you will need an angled USB extension cable that will fit inside the body to connect the mic and speaker.
When I first tried OpenAI's ChatGPT, I immediately thought it would be much better if I could simply talk to it and it would talk back - like a smarter version of an Amazon Alexa. So I decided to build one.
A brief video demo of Davinci - The ChatGPT AI Virtual Assistant
DaVinci - The ChatGPT AI virtual assistant is a voice-controlled and voice-response assistant that uses OpenAI’s artificial intelligence language model to assist with a wide range of tasks, such as answering questions, providing information, giving suggestions, telling jokes, writing stories and much more. In addition to providing responses verbally, DaVinci also provide written responses if your device is connected to a display.
Davinci - The ChatGPT AI virtual assistant is a python program that utilizes Picovoice Porcupine (and (PyAudio) to detect a wake word; Picovoice Cobra voice activity detection to determine when the user begins and finishes speaking their query; Picovoice Leopard to convert the spoken query to text; OpenAI ChatGPT as the artificial intelligence that responds to the query; Amazon Polly text to speech to convert the response into a natural-sounding human voice; and Pygame to play the audio.
DaVinci uses the most current ChatGPT API that was released on March 1, 2023. DaVinci will be updated to run on GPT4 as soon as that API is widely available to developers.
The Python code and detailed instructions for building and running DaVinci - The ChatGPT Virtual Assistant are on my GitHub site linked below. The instructions below are for running the DaVinci on a Raspberry Pi 4, but it can also be run on other systems with minimal modifications. The process for obtaining access keys for ChatGPT, Picovoice solutions and Amazon Polly is the same regardless of the system used.
DaVinci Body, Base Plate and Eye Sockets - Link to 3D Print Files
The main body holds the mic, the speaker and the LEDs. If you want DaVinci’s hair to be a different color than his body, just insert a change filament instruction at the approximate layer that the hair begins when you use your preferred slicer.
The base plate is the bottom part of the body. It has four pegs that the Raspberry Pi 4 will fit onto. You can permanently affix your Pi to the base by putting a little glue at each peg after the Pi is mounted. There is a hole on the back of the main body for the power supply plug, so be certain to arrange your Pi on the base in the correct direction. The base can be attached to the body with screws or glue.
The eye sockets fit into the holes for the eyes on the main body and the spherical LEDs will fit inside these sockets. You need to print two of them of course.
Schematics
Optional LED Wiring
DaVinci – The ChatGPT AI Virtual Assistant includes the ability to light up LEDs that help show the status of the program. If connected, the LEDs will light up when the program hears the wake word and will pulse after the user finishes their query and while the assistant provides its response. The proper wiring of the LEDs and resistors to the Raspberry Pi is shown on this diagram. The LEDs are controlled by GPIO pins 18 and 23. The use of the LEDs is optional and will not otherwise impact the functionality of DaVinci.
Code
Python Code for DaVinci
The python code and detailed instructions on how to install it and its dependencies on a Raspberry Pi 4 are here.
Comments