Raspberry Pi and ChatGPT bring AI conversations to your retro rotary phone

Who needs friends when you can make your own using AI? That seems to be what maker and developer Pollux Labs has pulled off with their latest Raspberry Pi project. Using our favorite SBC and a little help from ChatGPT, Pollux Labs upgraded a rotary phone to interact with those who call using speech recognition and text to speech for responses.

According to Pollux Labs, the project merges both vintage technology and the thrill of modern AI. To use the system, all you have to do is dial a specific number which then enables the speech to text function. You can talk with the ChatGPT AI in a full conversation that will be remembered so you can call again later and pick up where you left off.

The phone doesn't work as a regular phone but rather is monitored entirely by a Raspberry Pi. When the receiver is triggered, a dial tone sound is played like you would normally expect. However, it's up to you to call the AI program and engage with a futuristic experience that can only be accomplished with modern LLMs and a spark of ingenuity. So while you do retain much of the original experience using the handheld speaker and rotary dial, there's a huge element of modernity oozing out of the system.

Image 1 of 2

The project is driven by a Raspberry Pi 4B but you could get better performance using the latest Raspberry Pi 5. The Pi is housed inside of the old rotary phone along with a few components that bring the project together including a microphone that listens for audio which can then be parsed through the AI channel.

The software was put together by Pollux Labs just for this project and is primarily driven by a Python script that interacts with other services like the Whisper API from OpenAI to handle the ChatGPT interactions. It also maintains the conversation history and converts responses from ChatGPT into speech for the rotary phone to output.

If you want to get a closer look at this Raspberry Pi project in action, you can check it out in greater detail over at the project page shared to Hackster.