AI voice synthesis

[Mattias Larsen]

 

Wandered off to the odd side of youtube, found this quite interesting. Maybe someone here will too. Wonder if post can soon do ADR without the real actors. Or to repair lets say, just a word.

[Constantin]

Yes, we‘ve discussed this here before, but it’s worth bringing it up again from time to time. 
This technology, in my opinion, has the very real potential of destroying our profession. Once it really works well and easily, who will care about getting good sound on set when you can re-synthesize it all perfectly in post and even without the talent themselves present. 

[Fred Salles]

Fichtre diantre!!

 

Pretty impressive, and yes scary.

I do not think it will replace the actors in ADR soon, as the acting is not implemented yet (!!) but definitely might be used for repair in post.

[Olle Sjostrom]

I think it's not going to be long before this tech can be used in our business, surely. But it won't replace ADR entirely, because really, that's what it could potentially replace or apply to. Someone still has to record the voice on set. 

 

Think of it in terms of how CGI is used. You can do pretty much anything with CGI these days. Dead actor? No problem. But if you look at it long enough you're gonna see that it's an animated, dead, human. Uncanny valley. 

The same principle will probably apply to voice replacement. It will be a tool that will be useful for certain words here or there. That's what I think and hope.

 

And also, these examples are all studio recordings without a picture. AI will probably never be able to synthesize background noise and speech at the same time, and also sync to lips and emotion. But time will tell. By then though, digital actors will probably be more of an issue, but not for us as sound mixers.