OpenAI is slowly rolling out access to its ChatGPT Advanced Voice assistant to a “select group” of ChatGPT Plus subscribers. All paying users will be able to have a natural language conversation with the AI by the end of the year, but for now, it's only the lucky few.
At the end of last week, my account was flagged as one of those able to chat to an emotion-sensing, hyperactive, Yoda-impersonating artificial voice. After a full weekend with Advanced Voice, it is better and more expressive than the demos suggested.
One of the most notable features is the ability to interrupt the AI and have it immediately react to a change in direction. For example, I had it tell me a story about Paddington Train Station in London using a Yoda voice, then interrupted and switched to have it count to 100 quickly.
What stands out, though, is how ‘human-like’ Advanced Voice is compared to every other AI voice assistant I’ve tried. Speaking to it feels natural, and its voice reacts and adapts in tone and even speed to your voice as you talk to it.
I can see why OpenAI worries about people developing an emotional attachment to the AI voice. Combined with the natural language and knowledge of GPT-4o is a great experience.
Advanced Voice is a great storyteller
ChatGPT with GPT-4o is already very good at writing stories. Still, with the addition of speech-to-speech in Advanced Voice, it has also become a brilliant storyteller, able to adapt stories on the fly and even add multiple voices and energy levels.
I started asking it to tell me the tale of an AI that gains sentience, and it started well, sounding much like an audiobook. I then had it add elements like space travel and mathematical equations/real science. Then I made it ‘speak like vampire Yoda’, and it did exactly that. It sounds exactly like you’d imagine.
Next, I had it tell me a story about the first humans on Mars discovering something unexpected, including sound effects — which it did but sparingly.
I also had to ask it to be more dramatic in its reading, but it did so perfectly. It can also create a ‘make your own’ story where you steer the story. I asked it to have them find a human skeleton.
Advanced Voice as a City Guide
When I have to go to London for work, sometimes it is nice to look around and explore the area. My office is near Paddington Station, so I asked ChatGPT Advanced Voice to give me information about different sights and places.
This feature will become more useful as OpenAI integrates searchGPT and other live data features into Voice.
Even without live data, its training dataset is recent enough that it was able to tell me about the Paddington Bear statue, the history of the train station, and even details of its unique architecture.
Advanced Voice as a Personal Trainer
After decades of avoiding any form of exercise, I finally decided to get fit. I have a personal trainer and go to the gym regularly. Also swapping my cherryade addiction for water and eating more healthily generally.
I was suffering from an intense workout, so I asked Voice for some advice. It talked me through a stretching exercise, even counting me down from 10 to show how long I should hold a specific position or stretch.
It also talked me through different healthy recipe ideas. It motivated me while I was on the treadmill, offering regular encouraging phrases and adapting its tone and energy level to either a gentle persuasion or full-on drill sergeant.
Final thoughts
I don’t think I’ve even scratched the surface of what is possible with Advanced Voice yet. When I can access it by simply saying Hey, ChatGPT or tapping one button on my phone, it will also become significantly more useful — I hope Apple offers alternatives to Siri in the future.
When I was out walking and curious about a building I found myself asking Advanced Voice rather than typing into Google or ChatGPT.
When I first got access, I did all the silly things you’d expect, including having it try different voices, speak as Yoda, and count quickly. I also had it try to sing, speak in different languages and perform a short standup routine about space. I won’t be getting a Netflix special.
What I found though, as I used it more, is that it became a default way for me to look for information or interact with my phone. When in the supermarket, I used it to track what I was buying and even offer suggestions for alternative ingredients.
When I was out walking and curious about a building, I found myself asking Advanced Voice rather than typing into Google or ChatGPT.
Being so natural and responsive, with the ability to easily interrupt and change the conversation makes it a huge leap in computer interaction and one that has been due for decades. This is a leap on par with the mouse and touchscreen.
More from Tom's Guide
- OpenAI shares a new GPT-4o advanced voice demo — it can teach you a language
- ChatGPT Advanced Voice is out — 9 examples showing why you should be excited
- ChatGPT-4o Advanced Voice features — OpenAI just revealed when they’re coming
