01What Is the ChatGPT Voice Update?

The ChatGPT Voice Update is a new feature in the ChatGPT desktop app that lets you control your computer and run multiple AI agents using just your voice. It appears as a sticky pet on your desktop and can speak, listen, and coordinate tasks in the background.

This update is powered by GPT live, so it can understand your commands, reply naturally, and even take actions like opening apps or posting tweets. It is designed to be a hands-free way to manage your work without switching between tools.

02Main Features of the Voice Update

  • Sticky pet interface: The voice agent stays on your desktop like a small window. You can move it, bounce it like a basketball, or hide it.
  • Multiple voices: You can pick from different voices. The host prefers Maple. You can change the voice in the settings.
  • Screen context: The voice can see your screen to help with tasks. You can turn this on or off.
  • Voice chat hotkey: Press a button to start talking without opening the app.
  • Background work: The voice can run multiple AI agents in the background while you do other things.
  • Phone and remote access: You can use ChatGPT Voice on your phone to control your computer if it is switched on and running a remote session.

03How to Set Up ChatGPT Voice Update

  1. Open the ChatGPT desktop app on your computer.
  2. Go to the settings menu and click on the voice section.
  3. Choose your microphone from the list.
  4. Pick a voice you like. The host recommends Maple.
  5. Decide if you want the voice to have screen context (see your screen). Toggle it on or off.
  6. Set a voice chat hotkey if you want to start speaking quickly without opening the app.
  7. You can also hide the sticky pet if you do not want to see it on your desktop.

04How to Use the Voice Update: Examples from the Episode

During the episode, the host tested several tasks with ChatGPT Voice. Here is what it can do:

  • Post a tweet: The host said, “Can you post a tweet about the power of ChatGPT Voice?” The voice agent drafted a tweet, asked for approval, and then posted it on X. It also opened the browser to show the tweet live.
  • Fix audio settings: The host needed to record desktop sound. The voice agent opened the audio mixer and configured it correctly – something the host could not do before.
  • Journal in Notes: The host asked the voice to write a journal entry in the Notes app. It wrote a short note about the day’s activities.
  • Code in the background: The host asked the voice to improve the UI of a habit tracker in Codex. At first it failed, but after explaining again, it started coding in the background without interrupting the host.
  • Build a voice system: The voice agent began building an integrated voice system into the agent OS project, running multiple tasks in parallel.

You can also ask the voice to rearrange windows, open apps, or set timers – though the timer function did not work well in the test.

05Results, Limits, and Comparisons

The host said the ChatGPT Voice Update works well for many tasks but is not perfect. It posted a tweet successfully and fixed the audio mixer – both real-world wins. However, it failed on a few things: asking it to speak 50% faster did not work (it only increased speed by about 10%).

It also could not understand that the host’s camera was visible on screen, even after taking a screenshot.

When compared to building your own voice agent using the real-time API, the ChatGPT Voice Update is cheaper and smoother. The host said the API can get expensive and requires managing billing when credits run out. The update also runs multiple agents in the background, which a custom build would struggle to do.

06Tips, Limits, and Mistakes to Avoid

Here are some things to keep in mind from the episode:

  • There is a delay. The voice takes a moment to reply because it often uses thinking models. You may want to switch to a faster model like Instant if speed matters.
  • Some tasks fail. For example, setting a 10-second timer took longer than 10 seconds, so it was not ready when the host came back. Do not rely on the voice for time-sensitive jobs.
  • Train your brain. You need to remember the voice agent is there. The host said it takes muscle memory to start using it automatically.
  • Mute when needed. You can turn off the voice so it still works in the background without speaking out loud – useful if you are on a call.
  • Check the conversation log. You can click on the sticky pet to see everything the voice has done, including failed attempts.

07Who Should Use ChatGPT Voice Update and Why It Matters

The ChatGPT Voice Update is for anyone who wants to control their computer hands-free. It helps developers run coding tasks in the background while they focus on other work. Content creators can use it to post on social media, adjust settings, or take notes without typing.

It also matters because it is cheaper than building your own voice agent with the API. The host said it is a big step toward a Jarvis-like assistant, even though it is not that good yet. As the feature improves, it will become more reliable and faster.

08Background Parallel Agents: A Key Capability

One of the most useful parts of the ChatGPT Voice Update is that it can run multiple AI agents in the background at the same time. During the episode, the host started a coding task in Codex while also asking the voice to build a voice system in another project.

The voice agent acted like an orchestrator: it handled the requests, ran them in parallel tabs, and showed progress in sticky notes. You can even mute the voice so it works silently without interrupting you. This makes it possible to get real work done while the voice handles complex multi-step tasks.