James Earl Jones Signs Off on Using Recordings to Recreate Voice with AI | Hacker News Reader

James Earl Jones Signs Off on Using Recordings to Recreate Voice with AI

variety.com

39 pointscwan23 comments

The next step is musicians. We'll be seeing that very shortly.

It'll be possible to make new music featuring Whitney Houston, Elvis, Chester Bennington, et al.

On orthogonal planes, we'll see the rise of more virtual musicians with completely artificial voices (Gorillaz meets Hatsune Miku++) and voice correction for non-vocalists.

Source: I work on this tech and we'll be launching something very soon in this space.

Voice synthesis is also useful for people like Val Kilmer. The tracheotomy he underwent to treat his throat cancer wrecked his voice. Voice synthesis based on his own voice helps him sound like himself again:

https://www.sonantic.io/blog/helping-actor-val-kilmer-reclai...

Makes me wonder about copyright when it comes to voice synthesis via AI. Like legally speaking, did they "need" JEJ to sign off on this as opposed to just using his voice as training data?

In the near future, I imagine it'd be easy enough to take a snippet of audio and transfer a celebrity's voice to it, similar to what we can already do to images using StyleGAN or something similar.

Exciting times ahead!

I don't think they did, but celebrities are nothing if not PR machines, so I'm sure it was simpler to get this signed.

I think it was far less expensive for them to team up with Jones in the first place and create a PR buzz, rather than striking out on their own and risking a lawsuit by Jones or his estate, 100 years down the road. They fully realize this will soon be tested in the courts, and they don't want to pioneer that case law.

Looking forward to the JEJ Google Maps voice!

"I am altering the route. Pray I do not alter it any further. Or, tap 'no thanks' to stay on the current route."

Long long ago, in a Garmin far far away:

https://www.consumerreports.org/cro/news/2011/07/darth-vader...

https://fakeyou.com/tts/result/TR:qv34ck4p1y495bdxhr7q4k0acy...

I think the vocoder on this one is a bit off, but these are so trivial to make.

Off topic, but does this drive anyone else crazy? Why is changing my route opt out rather than opt in. As far as I can tell there's not even a setting to change that.

I'm with you. There is no setting. This drives me nuts on my motorcycle. I pick the nice scenic route, put the device into my pocket, (Bluetooth speakers in helmet) and then halfway through my very zen ride, this happens and I can do nothing but rage internally and try to remember to pull over and fix it, only for it to happen again 30 minutes later.

Google: you know I'm on a motorcycle. I see that on my location timeline page and marvel at the sophistication of technology. Please for the love of god if you won't give us a simple toggle for this behavior (I know you won't) just take the technology a tiny step farther and don't do this opt-out route changing thing if you detect I'm on a motorcycle.

How does it know you're on a motorcycle? I can imagine that it would be possible to detect fairly reliably with smartphone sensors (correlate curved ground path with gyro tilt), but it's astonishing and frightening that they would put in the effort to go that far.

Some 4-5 years ago someone analyzed an Android phone and showed they even know when you've exited a car.

I'll update this comment with a link when I find it.

I think you pretty much described the mechanism there. My friends and I have experimented with "riding like a car" and we can trick the algorithm for short distances if we really think about keeping the phone upright.

The keywords to learn more seem to be "activity recognition API" and I'm pretty sure they have been doing the motorcycle thing for the better part of a decade:

https://developers.google.com/location-context/activity-reco...

I wonder how much money was offered. Or even if Jones gets royalty/"performance" fees everytime a movie uses his voice. I imagine old actors would put that licensing agreement in their wills, and their kids and grandkids would get to get to enjoy the earnings.

That's the less interesting part IMO. The more interesting part is what happens when others make the same using public recordings... is this fair use?

Or training an AI using a James Earl Jones impersonator?

Tom Waits and Bette Midler have won lawsuits against companies who used voice impersonators in their ads.

Interesting times.

I wonder if their are stipulations around using the voice in advertisements or the ability to sell the voice to other companies.

Related, Hatsune Miku[1] was built with the voice from Fujita Saki[2], who was apparently offered a choice between either royalties or 50000 yen, and she chose 50000 yen immediately[3].

[1] https://en.wikipedia.org/wiki/Hatsune_Miku

[2] https://en.wikipedia.org/wiki/Saki_Fujita

[3] https://getnews.jp/archives/255685

That might have been a poor choice financially for Fujita Saki given that the current estimated number of Hatsune Miku songs is in the hundreds of thousands, but Hatsune Miku not being encumbered by royalties probably contributed to her popularity.

Star Wars: Let the past die. Kill it if you have to.

Also Star Wars: This.

I don't find the idea of movies/music produced this way appealing at all. Rather see something recast (or new stories told with new characters) versus seeing this done for the sake of safe nostalgia sales.

Perhaps if JEJ feels it is time to step away from the character upon reflection, Disney might do some reflecting as well.

Yeah, but JJ Abrams had no idea what he was doing so he was being ironical.

Recreation is an attempt to reshape the history.

History is the information residual form the past events onto various forms of recording, and projected into human brains, forming a collective residual image of the past that causes the nowadays events.

When history can be reshaped, the human collectives would lost its directional momentum because the causes are no longer entirely intact.

God bless us, when our children finally lost their roots...

Since the new Star Wars movies aren't canon, it's clear they'll to have to revisit older stories. Especially between Episode 3 and 4. So it's good insurance to get the voice, since Star Wars is and always has been a story about Anakin Skywalker.

When they take a stab at Episodes 7 - 9 again in a few years, they'll probably incorporate Anakin's force ghost more too. Although I suppose they don't need the voice. Might need it for flashback though.

Reply on news.ycombinator.com