Every new video showing off the capabilities of ChatGPT Voice leaves me more excited to try it out for myself, and the latest is no exception. In it we see the AI adopt a range of different character voices based on a simple voice prompt — perfect for storytelling.
It isn’t clear when the next version of ChatGPT Voice, also known as Omni Voice, will be available but rumors suggest the first users will get access later in the summer.
Unlike the current version of ChatGPT Voice, this new model is built using GPT-4o and is speech-to-speech natively, meaning it doesn’t first have to convert what you say into text.
This native voice modality is what allows the model to create different-sounding voices, express emotions and even detect signs of emotion in your voice while you speak to it.
What does the new ChatGPT demo show?
OpenAI have been gradually revealing the multitude of capabilities hidden inside GPT-4o's new voice mode. So far we've seen it translate conversations in real time, help with homework and even greet an audience at a French tech conference.
In the latest demo it opens with an OpenAI staffer giving instructions to the AI chatbot. He tells the AI he’s writing a story and wants to practice a couple of voices for different characters. One is a lion and ChatGPT puts on a gruff, majestic voice.
ChatGPT does an amazing job of the lion and is able to then quickly jump into the second character which is a "mouse that snuck into a cave."
What was really interesting was the way he was able to have the AI change the voice, telling it to make it "a little squeakier more like a tiny little mouse."
He then added other character ssuch as an owl that sounded wise, acting as an advisor to the lion and a villain character with evil laughter. ChatGPT gave a maniacal laugh! It created a rounded set of characters for use in the story.
Overall it did a great job and gives us an insight into how ChatGPT could potentially be used to act as Dungeon Master in a game of D&D or replace audiobooks with custom interactive stories generated on the fly.
When will ChatGPT Voice be available
OpenAI makes a point of clarifying that while voice mode is already available for all users in the ChatGPT app, the "new voice and vision capabilities with GPT-4o will be rolling out in the coming weeks."
Some users have started calling the new mode Omni Voice or GPT-4o voice. The features demonstrated in the new video are only available with GPT-4o voice and vision. Some users will be getting access in the next couple of months.
If you go into the iPhone or Android app and enter Voice mode you can see which version you are using by clicking the (i) icon in the top right. It should say new ChatGPT Voice "coming soon" if you're on the current version.
