Don’t watch Squid Game dubbed in English
When the Hollywood film industry was taking a big slice of the pie in the global market, voice dubbing for viewers who did not speak English played an important role. Today, we live in a globalized world where different types of content from different countries surprise people with their creativity and quality.
Korean series Squid Game, Netflix’s biggest debut hit, has captured viewers’ hearts, and now it’s time for AI to show off its capability of deepfake voice.
A famous movie director, who directed Thor: Ragnarok and the upcoming movie, Thor: Love and Thunder, tweeted on Twitter as follows.
Some agree with his opinion, some do not.
The original voice of the original actor is the best way to portray the movie as it is, as the voice not only reveals the unique identity of the character, but also the voice color and tone are really important when delivering movie lines. Moreover, sometimes the voice even provides nuances.
Without subtitles and dubbing, how are people who do not speak Korean supposed to understand the movie with all its nuances? The answer is simple.
With speech synthesis, we can hit a real voice changer, that is, the voice of the actual actors. To begin with, an AI must be trained with the actor’s actual voice data. Even if the actor’s native language is Korean, he should not worry about his English or Japanese accent. The AI takes care of that part and the viewers who do not speak Korean can watch the movie in their own native language and have a more immersive experience.
An AI voice generated from the actor’s actual voice would sound more natural than the dubbed voice.
Ethical concerns about voice deepfakes
If it’s that simple, would not there be an ethical issue? For example, could we use a fake person generator to create a virtual person and use the voice of a famous person so easily?
Yes, that could be the case, and we need to introduce a new set of rules along with the new technology. But we can at least tell if a person is real or not by looking at the picture or listening carefully to the voice. A clear voice with good quality is only possible if it is generated by good quality data.
For instance, the main character of Squid Game, Lee Jung Jae, has to record scripts with different emotional expressions in a studio to produce good quality of his doppelganger who speaks different languages.
If the voice you hear is of poor quality and sounds somewhat strange or awkward, it is reasonable to doubt whether or not it was produced by real data. In this case, it is likely that the creator of the voice is training the AI using random data he has been able to collect in various interviews and on TV shows.
However, we can see from the fact that many people still frequently search for “Natural text to speech” that the quality of AI voice is not yet that good. But things are moving forward, and soon you could be watching Netflix Original or Disney Plus Original content dubbed in different languages with the actual voices of original actors from various countries.
