Reddit leak shows that ChatGPT’s voice assistant has even more tricks up its sleeve

OpenAI’s ChatGPT voice assistant is clearly trying to advance voice commands.

In a sample of ChatGPT’s expected updated enhanced voice mode posted online, Redditor RozziTheCreator showed off what the new voice option sounds like and how sound effects can be incorporated into responses. Not only is the voice incredibly realistic, but well-timed thunderstorm sounds can be heard in the background, adding even more immersion.

We were already impressed when OpenAI unveiled its GPT-4o update earlier this year, showing off its AI assistant solving math problems and singing. But the company continued to impress us with demos uploaded to YouTube, where we can see ChatGPT inventing stories with multiple characters and different voices, and even two instances of GPT-4o interacting with each other. This latest leaked update with sound effects shows us that OpenAI’s ChatGPT has even more use cases than we initially gave it credit for.

A promising draft

When listening to the audio sample, the sound effects were somewhat similar to the sound of a real thunderstorm. Since this leak is likely due to an accidental release, we may get a more refined version once OpenAI eventually releases its enhanced speech mode. Still, it’s clear how some sound effects create something mysterious and dramatic to the random story you let ChatGPT create.

This may seem very simple at first, given all the AI ​​feats we’ve seen so far, but consider that the AI ​​chatbot was intelligent enough to create an appropriate sound byte and time it for maximum immersion, all while delivering its response in an incredibly human-like voice.

The added ability to create sound effects would be perfect for many things, like creating a bedtime story for your child or turning plain text into a more engaging and customized audiobook. One Redditor even suggested that it could replace your Dungeon Master during a Dungeons & Dragons campaign. There’s no shortage of possibilities here and there’s no doubt that OpenAI could improve this feature in the future.

Not yet ready for publication

While we got a sneak peek at ChatGPT’s enhanced voice mode thanks to this Reddit leak, it doesn’t look like it’s ready for an official release yet. Even the Redditor who stumbled upon the new mode said the voice assistant stopped working shortly after the sample audio ended. Furthermore, OpenAI recently announced that the launch of the enhanced voice mode had to be delayed as it needed more time for testing.

The company’s latest update was further complicated by legal issues; actress Scarlett Johansson threatened to sue OpenAI because the now-removed Sky option sounded too similar to the actress’ voice.

Whatever the case, the latest teaser of OpenAI’s new voice mode gives us a glimpse into the future of more advanced large language models powering AI chatbots and how they might complement voice input. Buckle up, because it gets even weirder.