Steve Jobs (& other voices) Reanimated via AI

Steve Jobs would have turned 68 a few days ago. His impact and legacy remain as powerful as ever, and many who were close to him (myself included) continue to mourn his loss. Recently, there have been a couple of attempts to reanimate Steve – at least his voice – using AI-assisted technologies. 

The first came out last Fall when a voice synthesis company based in Dubai, Play.ht, published a 20-minute fictional podcast interview between Steve Jobs and Joe Rogan using semi-realistic voice cloning technology based on the AI learning from the voices of both men through a broad range of samples. It’s a little surreal. In this fake interview, Steve’s voice in particular is at times choppy and clunky. That said, the advancement of deep learning technology has allowed AI models to replicate distinctive voices with an ever increasing degree of accuracy which is both exciting and terrifying.

The second came out about 10 days or so ago, through an Apple Alum, Michael Darius, of .fm/imagine. Michael is a top notch designer who wrote many of Steve’s demo scripts for his presentations, and who played a key role at Apple where he was responsible for designing the earliest versions of some of the company’s most core services including iTunes, iLife, GarageBand and iCloud. Michael is committed to preserving and protecting Steve’s legacy as a creative visionary, and in tandem to revitalize the design community that Apple has been at the center of for decades.

This demo that Michael and his team created which appears to showcase Steve presenting ChatGPT over 20 years ago is pretty impressive. Again, this was done using AI technology trained on Steve’s voice enhanced by Michael’s up close experience with Steve’s stage presence, nuances and tonality . It’s a brief demo (less than 4 min in length); check it out here (click on the play arrow in the center visual). You’ll want to experience it!

I have to say with all due respect, the demo does a good job capturing Steve’s snarkiness. When I first heard it and didn’t know it was a fake, I actually was gobsmacked at the thought of him presenting ChatGPT! Until I learned otherwise, it had me thinking that maybe Steve had indeed come up with ChatGPT over 20 years ago. I leave it to your interpretation and imagination to see what you think.

Finally, in the context of the topic of cloning famous voices through AI, I’d be remiss if I didn’t mention Sonantic, based in the UK, who helped actor Val Kilmer make a memorable appearance in Top Gun: Maverick. The team at Sonantic were able to recreate Kilmer’s speaking voice by using a “voice engine” to train the AI voice model through hours of audio. Sonantic was since acquired by Spotify.

The legal ramifications of voice cloning through AI remain to be seen (especially if there’s commercial intent), but these early attempts are pretty fascinating – and this space will definitely be one to watch.