If you’ve ever wanted a personal AI assistant like Scarlett Johansson‘s character in the movie Her, that reality might now be a little bit closer.

OpenAI has unveiled a new version of its AI chatbot called ChatGPT-4o (the ‘o’ is for ‘omni’) which appears to be much more human-like than earlier versions.

According to OpenAI, the new model can respond to any combination of audio, text, or images – and respond with a mixture of the three.

And, just like in Her, the newly upgraded voice mode can chat with human-like responsiveness and even read users’ emotions.

Speaking at the new bot’s launch event, OpenAI’s chief technology officer, Mira Murati said: ‘We’re looking at the the future of interaction between ourselves and the machines.’

OpenAI has launched a new version of its AI chatbot, creating a new incredibly human-like personal AI assistant just like the movie Her

On social media, tech fans were quick to point out the obvious parallel between OpenAI’s creation and this 2013 sci-fi romance novel, with one user jokingly congratulating the company, saying: ‘Finally, you did it’

During the live-streamed launch, OpenAI showed off some of the model’s impressive capabilities.

Mr Murati and OpenAI engineers asked the AI to change up its presentation of a bedtime story, use a robotic voice, and even sing.

At one point a researcher asked the AI to look at their face and judge what their emotion might be.

And, after a small technical issue, the voice assistant replied: ‘Happy and cheerful with a big smile and maybe even a touch of excitement.

‘Whatever’s going on, it seems like you’re in a great mood, care to share the source of those good vibes?’

In other videos posted by OpenAI, the new voice mode teached maths, harmonises with another voice assistant, and even helped a blind man hail a cab while navigating London.

The comparison to the movie Her was not lost on tech fans, with one social media user making a spoof poster featuring OpenAI CEO Sam Altman’s face (right) in place of Joaquin Phoenix

OpenAI CEO is also open about the inspiration for the voice assistant. Altman even wrote in X, formerly Twitter, a simple message just saying: ‘her’

And with its cheerfully expressive female voice, the comparison to Scarlett Johansson’s AI assistant character in Her is almost unavoidable.

Nor does this comparison seem entirely unintentional with OpenAI CEO Sam Altman posting a message to X, formerly Twitter, simply saying: ‘her’.

In a blog post about the launch, Mr Altman also wrote: ‘The new voice (and video) mode is the best computer interface I’ve ever used.

‘It feels like AI from the movies; and it’s still a bit surprising to me that it’s real.’

Part of the reason that this AI assistant feels so life-like is due to its lighting quick reaction times.

ChatGPT-4o can respond to audio inputs in as little as 232 milliseconds, with an average of 320 milliseconds – within the range of normal human responses.

OpenAI under CEO Sam Altman has risen to rival the established tech giants. The timing of this announcement also comes one day before Google’s biggest event of the year, seen by many as a challenge to the company’s rival

Previous versions of ChatGPT did include a voice mode but this was slow and lacked expression in its responses.

This is because it actually used three separate AI models: one to turn the audio to text, one to respond to the text, and another to turn that prompt back into speech.

This meant that the AI couldn’t directly observe tone, multiple speakers, or background noises, and couldn’t output laughter, singing, or express emotion.

However, version 4o uses one model to handle the entire input and output process, making it much more responsive to subtleties of tone and emotion.

OpenAI said it was making the new, enhanced model available to all ChatGPT users, even those who do not have a paid subscription with the firm as part of its ‘mission’ to ensure AI technology was ‘accessible and beneficial to everyone’.

However, the voice assistant capabilities will be launched in an alpha model in the coming weeks and will be available only to ChatGPT Plus subscribers during a wider rollout.

The new voice mode can respond within as little as 232 milliseconds, with an average of 320 milliseconds. OpenAI claims in its demonstration that this is close to human response times

At a time when the scrutiny around AI tools continues to grow, OpenAI said the new model had been extensively safety tested, using independent experts, and has ‘safety built-in by design’.

OpenAI admits that the AI’s new voice capabilities creates ‘a variety of novel risks’ and notes that only preset voices will be available at launch.

While the company doesn’t specify, these ‘risks’ are likely to refer to deepfake phishing scams or other forms of audio manipulation.

In these attacks, scammers use AI to generate realistic audio of loved ones asking for money or claiming to be in distress.

And while OpenAI insists that the model will ‘abide by our existing safety policies’ it is easy to see how this advanced tool could be used in this sort of attack.

GPT-4o as tested by @BeMyEyes: pic.twitter.com/WeAoVmxUFH

— Greg Brockman (@gdb) May 14, 2024

In one impressive demonstration, ChatGPT helps a blind man (pictured) hail a cab while navigating London by describing what the phone camera can see

The launch of this new model also comes at a particularly key time for OpenAI.

Reports in Bloomberg suggest that OpenAI may be reaching the end of talks with Apple to provide a new generation of AI-powered assistants.

Given Siri’s notoriously unreliable responses, a phone assistant with the capabilities of ChatGPT-4o is a highly appealing prospect.

OpenAI’s announcement also fell on the day before Google’s I/O event, at which the tech giant is expected to unveil its own AI products and features.

While it was rumoured that OpenAI would be releasing a search product to rival Google, this turned out to be false.

However, the launch has still been interpreted by many as a direct challenge to the company’s more established rival.