Until a little while ago, “AI dubbing” was just, well, “AI dubbing.” You know, the thing that would replace “traditional dubbing”, like, next year. Or maybe the year after. Or – soon. You would need a human in the loop to make it perfect. Or a copilot. Or a “human first”, which actually means, humans get to be the first to look at what AI produced and to fix it, like, add that famous “cultural nuance”. The 20 or 10 or 2 percent that the machine can’t, simply cannot, produce.
Maybe that all of this began to sound a bit scary. People were perhaps wondering how to identify what exactly is hiding in those missing percentage points. Is there an algorithm that can function as a nuance-detector? Is the uncanny valley bridgeable? What will happen to the industry? Which kind of language service provider will survive? Maybe only those who can provide “boutique” dubs while the rest is done not by LSP’s but by the producers of entertainment right at the source? What is going to happen to the people that once used to write and translate, direct and act? What do the audiences say?
No one really likes to say, hey, “AI dubbing” is here and it will replace all of you (well, some people do but they are rare). It sounds much better if you say, hey, “AI dubbing” is here and it’s a useful tool. It will help you.
Okay, maybe not you, the creative.
Synthetic voices will not help voice actors, it will replace them. I know that they are advertised as a new business model for actors – license your voice and earn money while not actually acting – but honestly now. Actors are actors because guess what? Because they want to act. Plus, I see none of the licensing models offering enough money to really make a living. And for how long?
For writing dubbing scripts, I haven’t found AI helpful. The only AI tool that I could possibly be using is ChatGPT, and although I know colleagues who use it as a sparring partner to discuss ideas – looking for alternatives and synonyms, or writing background dialogue – it doesn’t work with my creative process. I’m also not sure about discussing what’s essentially classified material with a program that uses its input for training.
I have, on occasion, checked a translation that I didn’t trust against DeepL, I do admit that. If the translation seemed really fishy, I took my issues up with the translator themselves, so I didn’t need DeepL for very long.
Other than that, no AI that I have ever seen addresses any of the issues that I might encounter while writing.
I can think of some. I have wished on occasion that they could pry the odd lip open in order to make something like “marry me” look more like “heirate mich.” And I know that I am not alone. So, give me deepfake tool and the artistic license to use it, and I will, if I think that a change in the image would be in the interest of maintaining the suspension of disbelief that is necessary for a dub to work.
Every dubbing scriptwriter I know has cursed to high heaven when a character suddenly appears that they know has appeared before, but did they address their best friend’s mother formally or informally? Damned if you didn’t spend hours looking through old as-rec scripts only to find out that your colleague from season 2 has avoided the issue altogether.
Every dubbing scriptwriter I know hates writing recaps. We spend useless hours digging through perhaps several seasons worth or as-rec scripts just to find the excerpts. AI to the rescue please!
So, if AI were really used as a tool in the dubbing process, it would be used by creatives to make things faster AND better. It would be used by dubbing companies to make, say, their contract management less cumbersome, or assist in booking talent. And it would be good if dubbing companies invested in this.
I have said elsewhere that I don’t believe that the term “AI dubbing” is useful. There is only dubbing, and the dubbing process utilizes any number of different workflows, with more or less degree of automation.
What does exist are “AI dubbing companies”. And they don’t use “AI dubbing” as a tool. They ARE the tool used by content providers in order to localize their product. For content providers, “AI dubbing” (provided it really does end up being faster and cheaper than “traditional dubbing”, never mind better) is the most useful tool since the invention of the internet. Whether individual creatives fit into this process, and how that spark of creativity that should be in the center of the adaptation process will survive, is extremely uncertain. And this is why the sentence, “AI dubbing is a useful tool” doesn’t sound at all comforting to me.
