Val Kilmer and the Future of Podcasts.

A few months ago, an interesting video was uploaded to YouTube, which did not receive much attention. Val Kilmer, the movie star most remembered for his role in 'Top Gun', was diagnosed with throat cancer and as a result, lost his voice.?

This year he got it back.

Sonantic engineers collected samples of Kilmer's voice from the films in which he played, and used them to train an #artificialintelligence system that produces "Deep Fake" of his voice.?

Meaning: The system takes words written by Kilmer, and turns them into speech in his own voice.?

Judging by the video (see the first comment) the quality of speech is truly stunning, including an (almost) precise intonation and appropriate to the emotion expressed by the text. If I did not know it was Deep Fake, I don't think I would have guessed it myself.

Video Deep Fakes reached a peak a 1-2 years ago, but this is the first time I have encountered such a high-level Audio Fake.

The success of Sonantic has made me reflect on the possible implications of this technology for the world of podcasts. Sure, this technology will have interesting implications outside of the podcast world as well, but I might expand on that on another occasion.

Assuming (very likely) that this technology will soon be available to anyone, it may make human speech another "tool" in the podcast producers' toolbox - similar to the other tools we use today such as music, effects, overtures and closures. What does this actually mean?

(1) No need to bring guests to the studio, or record an interview at all. If I have enough voice samples of the interviewee to train the neural network to a speech level (for example, recordings from YouTube videos, etc.) - I can "interview" him or her via email or WhatsApp, and turn the text into fluent speech. This will revolutionize the production process of a podcast.

(2) Podcasts will be able to produce many more programs at the same time. There will be no problem, in principle, to train a system so that you can emulate a show host voice perfectly. In such a situation, you can let someone else write the texts - and the system will read them out loud... In fact, it will be possible to continue publishing new episodes of a podcast even many years after the host has long passed away!?

(3) The voice of a podcaster will become a 'digital asset' that can be easily sold or rented, for passive income. Do you want Joe Rogan to host a podcast for you? No problem - for a fee you could buy a license to use their voice.

(4) Lastly, imagine translating your podcast into other languages with simplicity and ease. Once a translator has finished working, press a button and you have the episode in any language he has translated into, in a perfect accent. Imagine your podcast translated in some common languages, say English, Spanish, Chinese, Arabic with the associated increase in market volume from advertising, relative to production cost.

On the other hand, this development is likely to challenge beginner podcasters, as the competition for the audience's attention will become much tougher. If today there are about two million podcasts in the US alone, the possibility of creating a high quality and relatively easy 'fake' podcast could flood the market with millions of new podcasts, presented by successful and famous podcasts hosts who 'rented' their voice for a fee.

There is also the issue of authenticity. Is the Deep Fake podcast a "real" podcast? Would you be willing to listen to Joe Rogan's show, if you knew that it's “not really him??

These are fascinating questions, which will surely occupy the podcasts world in the not so distant future ??

From Hebrew: https://www.facebook.com/groups/rotmil/permalink/2704886179814881/