Opinion | Hank Azaria’s “Simpsons” voice is not completely replicated by AI

I think soon, artificial intelligence can replicate the sounds of over 100 voices they’ve created for the characters in “The Simpsons” for almost 40 years. Thinking about it makes me sad. Needless to say, it seems totally wrong to steal my likeness, sounds, or anything else.

In my case, AI has access to 36 years of MOE, a permanently dissatisfied bartender. He appeared in almost every episode of “The Simpsons.” He was terrified with love, his head raged, mostly in a state of bitter hatred. I’ve laughed as Moe in dozens of ways ever. I probably sighed as Moe 100 times. When it comes to AI training, it does a lot of work.

But the voice is more than just a sound. And I would like to assume that even if the AI ​​version of MOE, SNEKI, or Chief Wigum sounds like my voice, something is still missing. Who I am depends on creating a voice. How can a computer remind you everything?

The only misconception about voice acting is that you only need a voice. But our bodies and souls are involved in achieving proper reliability. I’m recording vocals when I first saw Dan Castellaneta playing Homer and Dan Castellaneta playing Harry Shearer playing Mr. Burns and many other characters, so I’m just stupid to see them. I was almost embarrassed to see it. They were flying around and giving everyone a perfect performance – just a microphone. I was 23 years old. It took me a while to get the courage to do that.

If the character is running, it’s as easy as running in place. If your character is crying, you cultivate real tears, real emotions. Many of my characters were thrown punches or punched in the face. If your character is talking while throwing a punch, it’s hard to fake unless you actually throw a punch. You may get props if they help you get into the reality of the scene. I played a character who was tasting cigars, so while I was talking I stuck highlighters in my mouth.

It has always been interesting to see major movie stars and great actors who have not spoken out much over the years, and to see them come to document them together. They don’t know that they were unable to do it at first after climbing their heads. When they realised it, they were great. I remember Mandy Patinkin and Anne Bancroft coming in and realising it. Mick Jagger isn’t shy on stage, but he had to make the journey. He ultimately got you to have to be totally committed, like you do with any performance.

Another thing we do in “The Simpsons” is improvisation. When you play and interact, there is interruptions and natural interactions – you don’t just recite things by line. It’s hard to imagine a computer can mimic that rhythm.

Over the years I created the voice of Professor Flink, a comic book man. They were created in all sorts of ways, including imitating celebrities, friends, and family.

When I arrived at the audition for “The Simpsons,” I went in and gave the impression of a young Al Pacino. At the time I was playing a drug dealer in a play, talking like a young Al Pacino in that role. When I did it for an audition for the “Simpsons,” he said, “We like that voice, but we want you to make some gravel.” You take my young Al Pacino version and add gravel to it and you make Mo a bartender.

Chief Wiggm is an imitation of Mel Blanc, who gives the exaggerated impression of Edward G. Robinson. I grew up hearing that. One of the most satisfying things about being a Simpson is that for the kids who grew up on the show the way Mel Blanc and Bugs Bunny did to me, it seems to mean staying with me. they. Can AI do that for people?

Anyone who is imitation or makes the impression of a voice is already like a strange version of AI. You can save these voices, remember them deeply and recreate them. But for Chief Wiggm, I am not a straightforward imitation of Edward G. Robinson. A computer might be able to do that. I’m doing the weird impersonation.

The voice I created when I played Agador, a butler who tried out shoes in the 1996 film The Birdcage, came from my memories of my childhood. There were two voices that I had decided on for the character. One was tougher, like the Puerto Ricans I grew up listening to in my neighborhood in Queens. The other sounded like my maternal grandmother.

In my family, we were Sephardic Jews from a bilingual Spanish and English family. My grandmother spoke five languages ​​and when she spoke English she had a Hispanic accent. She was also very loving, sweet and feminine. I’m not the most macho guy in the world, but my character was very motherly to the other characters in the film. I had nothing to do with it so I began to imagine what my grandmother would do, and it all clicked for me. So it didn’t just sound like her. It was her mentality and affection that took place in creating Agador’s voice.

How does the lack of humanity sound if AI tries to replicate one of my voices? How much is the difference? I honestly don’t know, but I think it’s enough to just realize something is off, just like something is wrong with a sub-movie or a TV show. When the expo is clunky, or when the dialogue is bad, or when the character says something outside the character – why would he say it if he was afraid? Why did she publish such a backstory for her? et cetera.

It’s a feeling that what we see is not authentic, and you don’t need to pay attention to it. Faithfulness is acquired through craftsmanship, excellent storytelling and excellent performance, excellent cinematography, excellent director, excellent script and excellent music.

There are only small enough voices generated by AI to make you think that something is missing. Just as it appears that the faces produced in the video lack the elements that make them believe, it’s just not persuasive or interesting.

Or it could depend on the episode. Great writers don’t slap it from the park every time. They provide you with great scripts, medium scripts, non-OK scripts. Perhaps that will be the case with AI. We also recognize that in our distracted times, people may not be able to keep up with the differences.

There may be several aspects to the performance that AI can enhance. I know that certain lines need to laugh, but I don’t know how to get it, so I’ll try something else. Create a list of 8 or 9 ways to try it out. I do crazy take, happy take, sad take, deadpan take, aggressive, something that really feels in my mind. It’s hard to tell which one works, but you can always communicate it in your editing.

AI models may not know what is interesting or what timing is, but they could make a million different take. And it might be said to do them the same as me – and it might be pretty convincing.

So, if I’m honest, I’m a little worried. This is my job. This is what I want to do and I don’t need to stop doing it. The traditional wisdom in Hollywood is that the technology to make your face completely human is five years away. I’m worried that the audio equivalent will also be coming.

If AI takes over, there may be some benefits. I miss Melblanc’s old bug bunny performance. I will never get them again. But perhaps with AI you can get more of them. That might work especially well if someone like me who is intimately familiar with the subtlety of the characters can help them recreate what Bugs was doing by essentially directing AI. yeah.

I think we still need someone who knows what it needs to do in his heart, his heart and his soul. AI can make sounds, but it still requires people to make performances. Computers understand emotions on their own, what is moving, what is interesting? Now we’re in the process of science fiction. To do this, I think AI must be alive.