When she’s not being Black Widow, one of Scarlett Johansson’s most beloved performances in recent years was as the voice for the AI Samantha in 2013’s Her. Now, rather suitably, her voice has been synthesised using AI to make a video.

The channel behind this is Stable Voice. Over the past few months it has made a name for itself by teaching computers to talk like celebrities, and then making them “read” funny scripts.

They caught waves with their “A Stable Genius Explains” series of videos in which a fake Donald Trump talks on a number of subjects. You may think this was used for some malicious speeches, but it’s all in good fun as the robot US president explains topics such as bone spurs, flags and coal.

The full Stable Genius Explains series can be watched on YouTube, but today we’re looking at what the channel did to Johansson voice. Stable Voices has chosen to replicate part of the scene where Samantha is installed (here it is for reference) before it takes a weird turn.

To be completely honest this isn’t the best work from the channel we’ve seen, as Johansson sounds like she’s suffering from a serious head injury and / or is extremely drunk. It’s still impressive and it’s definitely recognisable as the celebrity, but it is clear that something is amiss here.

If you’re not convinced by Scarlett Johansson or Donald Trump, Stable Voices has also tackled Alan Rickman as Severus Snape, Brian Cranston as Walter White, Jeff Goldblum as an Area 51 guide, and many more.