Tony KipkemboiAI Engineer & Content Creator

I used my voice with ChatGPT 🫣 #shorts

5 Mar 2023 · 1 min

Watch on YouTube

1:01

In this short video, the presenter demonstrates a Python application that combines OpenAI's Whisper speech‑to‑text API and the ChatGPT API to enable a two‑way voice conversation with the model. The app records the user's voice, sends the audio to Whisper for transcription, forwards the resulting text to ChatGPT, receives a textual response, and then uses a text‑to‑speech component to read the answer aloud in real time. The demo shows the full interaction loop by asking the model "Where are you from?" and playing the spoken reply. The presenter emphasizes that the entire solution is built solely with Python and requires no additional languages or platforms.

Chapters

I can talk to ChatGPT. I can ask it a question and it can answer it back in a voice. And here's how. A few days ago, OpenAI released 2 APIs. One is ChatGPT API and the other one is Whisper, which allows you to turn audio into text. I went ahead and created a Python app that allows me to use the 2 APIs to talk to ChatGPT and for it to talk to me. This is the app I built and you can build this with just Python, no other language required. App, I can record my voice, which then gets transcribed to text using Whisper API. Then the text gets fed to ChatGPT API, which generates this response right here. And then I also have a piece of the code that reads out this response back to me in real time. Now let's give it a try. Where are you from?
As an AI language model, I was not born in a physical location. However, I am housed on servers and workstations maintained by OpenAI, which has headquarters in San Francisco.