I’ve created and open-sourced VoxGPT, a web app that uses OpenAI Whisper to provide a conversational voice interface for GPT-4 and GPT-3.5. It’s accessible from any modern browser, including mobile browsers. It enables users to verbally communicate with the latest OpenAI completion models.
Here’s the repo:
And here’s a quick demo video:
Pretty cool project. I’ve got a different approach to doing the same thing with
Mine’s a browser extension which requires you to be logged in to ChatGPT.
You’re using your own key for the demo? I guess that works as long as it’s not too popular 
This is really cool and exactly what I was looking for. Ideally there would be a text box added where the text inputs and outputs could be seen and copied.
Which temperature settings did you use? The answers seen quite factual.
Yes, using my key. So far has been very inexpensive, but I would likely take the demo down or have the user input their API key if it becomes more popular.
The temperature defaults to 1, as documented here: OpenAI API
Excelente idea, y como componente Web , se puede implementar como asistente en cualquier pagina Web.
I’ve added a number of improvements so it’s worth checking it out again: