Codex Voice: Build Your Own JARVIS!

Julian Goldie (0:00)

This is a powerful new voice update inside Codex, and it turns your computer into Jarvis. You use Torque and it builds websites, apps, and tools for you whilst you watch it happen inside your terminal. Today, I'm going to show you exactly how it works, and then I'll show you something even cooler. I use it to build my voice agent that can control my whole computer. It can open things, run things, do work for me, or with my voice. And it took minutes to set up. And later in this video, I'll show you the one command you need to know to make this all run smoothly. If you stick me to the end, I promise you'll walk away able to build things just by talking inside your terminal. Let's get into it. There's a brand new update inside OpenAI Terminal now. So if you're using Codex inside your terminal, you can basically use your agents to code for you using your voice. And I'll show you exactly how that works in a step. So basically what you can do here is you can make sure, number one, you have Codex installed locally, and make sure you signed in to your Chat Cheapity account. I think this will probably be available just for the plus and the pro users. And then what you're going to do from here is you're going to open up your terminal. So we actually did this example earlier, I actually built out and coded this whole page for me using my voice, which was pretty fun and easy to do. So what we did inside the terminal, as you can see right here, is we said, okay, I want you to build out a new website for the AI Profile Boardroom using voice. And then it just went off, coded out for me, and it can actually look into it. So I said, what do you think about the design so far? And it can actually give you feedback. So it's essentially turning Codex and ChatGPT into Jarvis. Now, you could previously do this inside the Codex app directly. I actually think this is a lot smoother and easier because it's just so simplified if you're using the terminal. It feels a lot nicer to use, to be honest. So there's an example on the demo video where basically, the coder is managing their projects across like a burrito app or something like that. There's no real documentation on how to use it. But basically, if you want to get started with this, just make sure number one, you've updated it. So you can just make sure you have Codex fully updated. And then once you've done that, type voice like so.

Hey, can you build out a app for fitness for me, please?

SPEAKER_2 (2:12)

Sure thing. Let's get it wrong.

Julian Goldie (2:14)

Stop. Carry on.

SPEAKER_2 (2:19)

One sec. I'm on it.

Julian Goldie (2:20)

All right, cool. One second. Can you actually like see my screen right now?

SPEAKER_2 (2:26)

Checking that now.

Julian Goldie (2:28)

So you can see here is actually pretty slow when it comes to seeing the screen. So I wouldn't recommend that. But it is decent for like coding projects. And if you want to stop this, you can just type forward slash stop like so. And then it will actually stop the voice project as you go along. So it's pretty easy to like switch in and out of it. Then if you want to interrupt it, because it seems to be super slow, then we can go back here.

And we'll switch back to voice. Hey, you there?

Yep.

SPEAKER_2 (2:56)

I'm here. What's up?

Julian Goldie (2:58)

How fast do you actually run?

SPEAKER_2 (3:01)

Pretty quick for chat stuff, but it depends on the task.

Simple things can be answered almost instantly, while bigger jobs take longer. All about the workload.

Julian Goldie (3:11)

Can I change your voice? Can you change your voice?

SPEAKER_2 (3:16)

I can't change it from here. That's controlled by your app or device settings. If there's a voice or audio option, it could be there.

Julian Goldie (3:24)

Okay. Can you build out an app where I can actually use your voice from Codex inside, like, just a simple screen?

SPEAKER_2 (3:34)

Let me check that.

Julian Goldie (3:35)

So you can actually switch off voice by just typing in forward slash voice stop. And then it ran for about five minutes to build this out. And now we've got the voice agent here. So literally built like, you know, a voice agent in a couple of clicks. And if we click on start conversation here, and let's just test out. Sometimes it can build something like this. And then, you know, you have to go back and forth and wrestle with it a little bit. So let's just click on start conversation. We have to allow the mic.