LocWorld52: Farbod Mansorian on the Rise of Multilingual YouTube

Episode 228 November 07, 2024 00:26:54
LocWorld52: Farbod Mansorian on the Rise of Multilingual YouTube
Localization Today
LocWorld52: Farbod Mansorian on the Rise of Multilingual YouTube

Nov 07 2024 | 00:26:54

/

Hosted By

Eddie Arrieta

Show Notes

A conversation with Farbod Mansorian, the founder and CEO of Unilingo. After producing over 5,000 videos and reaching 50 billion views across 14 languages, Farbod shares his vision for YouTube dubbing, insights on content localization, and his thoughts on the future of the language industry.

View Full Transcript

Episode Transcript

[00:00:03] Speaker A: The following is our conversation with Farbod Manzorian, the founder and CEO of Unilingo, after 5,000 videos and 50 billion views in 14 different languages, talks to us about what dubbing for YouTubers means to him and what he thinks is the future of the language industry, especially in dubbing. My name is Eddie Arrieta. I'm the CEO of Multilingual magazine. Enjoy. [00:00:41] Speaker B: I first generation immigrants I moved to, you know, one day I was in my home country, next day I was in Canada. And I learned to adapt very quickly. I did not speak a word of English prior to the age of 17. I moved there with my father. He then went back and I had to figure out essentially life that built my character and gave my work ethics. I studied mechanical engineering to go and work on in the automotive industry. I ended up making a startup in the hardware space. We sold the assets. It was these power banks that would be dispensed through a station, placed at high foot traffic areas targeted at tourists. We sold the assets and then later it became larger or bigger company where you go to a live nation concert and you can charge your phone on the go. That was my first startup. Then I started two more companies which failed. And around 2017 I got accepted into effectively the Y Combinator of Canada, which is called Next Canada. And I was toying around with different ideas and I was at the time translating TED Talks for family and friends back home. And it was a hobby of mine. And it was a hobby because I learned English and all of a sudden I had access to world's greatest information simply because I knew this language. It's almost, it's like in the 1700s, we had to learn Latin to be able to consume knowledge. In today's day and age, if you want to, you know, consume information, we most probably have to be able to speak English. And that will likely change very soon. So I was translating videos. Then I, as an experiment, I took one of these TED talks and I dubbed it in Spanish. And the reason was because I learned that the Spanish fans on YouTube, which we were translating and posting it on YouTube. Ted Talks has a YouTube channel. So we dubbed the video in Spanish. We uploaded and I moved to, I go to San Francisco, San Bruno, where the YouTube headquarters are, and I pitched 300 YouTube engineers to join me on my new startup, which is translation for YouTube creators. And I got 300 nos with a broken heart going back to my Airbnb. I looked at my phone and the TED Talk in Spanish had a million views. And I took that As a sign to lean in and start a company around localization for content creators. But not just localized, but also post and distribute it for them. So create a localized version of their main YouTube channel. I moved to Argentina to try to get as far away from the cold weather, maybe that's. And I built a dubbing team. Not knowing anything about dubbing, we started to work with this content creator's name is Veritasium. He is the second largest science channel on YouTube. We grew his channel from 0 to 3 million or 2.5 million subscribers. Hundreds of millions of views, hundreds of thousands of dollars in additional AdSense revenue. And fast forward to the day. We have dubbed and distributed 7,000 YouTube videos for some of the largest creators in the world. We've generated111.10 billion. 10, 11 billion views and $32 million in additional AdSense revenue. And we work with creators like Mr. Beast, PewDiePie, Dude, Perfect. Also Steve O. And now we are more focused on implementation of AGI to bring the cost of dubbing down so that it is more accessible for everyone. Sort of started in 201312 with the founder of Watchmojo, which does the top 10 videos category. And he created about 14 YouTube channels in different languages. And he was very successful. So Watchmojo in Espanol had 8 million subscribers. So he started that movement in 2012, 2013. Then came a group of kids channels like Nas, for example, like Nastya Ryan's World, where they started to dub. But it made sense because it was kids content and they were getting insane amount of engagement. But none of the mainstream creators were doing it. And it was primarily because there was not a platform where I could just go push a button. And then things would just happen for me. Most of the time they would have to go through a dubbing studio. And the dubbing studio would say, oh, if you have X amount of talent in the video, I will charge you this per minute and then I will have to charge you extra amount of dollars if you want this something else. And creators just want to create content. So knowing this, I approached creators and I said, hey, just sign this contract and push this button and I will dub and distribute your videos. And I think that's what clicked. So in 2017, we came around and we started dubbing YouTube videos. And we really, especially around 2020, with Mr. Beast in Espanol, we shot or dubbing for YouTube creators shot into the spotlight. Because I took Mr. Beast and Espanol from 0 to 1 million subscribers in 28 days. It was the third fastest growing YouTube channel. In that year, we. One of some of the things we did was we hired the best voice actors, best talent. But we also partnered with regional creators in Mexico and Colombia and Chile and Argentina to promote his translated content to their audience. And on top of that, we responded to comments. We responded to about 20,000 comments. We engaged with the fans, and then YouTube learned that they can implement the multi language audio feature, much like Netflix. So we uploaded a few Audio tracks to MrBeast. The first audio track we uploaded was to get in this top for 10, get this in this top full of snakes for $10,000. And we dubbed it in Spanish. We uploaded it as a multi language audio feature, but it failed massively. And then I said, jimmy, the multi language audio feature isn't a fully mature product. Let's create a separate channel like I already have for other creators and let's see where it goes. And that was a massive success, I think. In 2018, I went to Vid Summit, which is one of the larger gatherings of YouTubers, and I had this one creator who said, youtubers will never want to pay you $50 per minute to dub their videos. It's too expensive. And. Or he would all. And on top of that, he would say, youtubers just don't understand the idea. Well, last month I was looking through our database on our website and we received 49,000 submissions from YouTube creators to work with us. Now, granted, we are expensive for most of them and so dubbing is very cost prohibitive. This is, in my opinion, I think, less than. Arguably 1% of all shows are being localized because it's just so expensive. And let's talk about subtitles versus dub for a moment. Early on, I ran a poll on YouTube where I asked a simple question. On one of our videos, do you prefer subtitles or dub? We got 120,000 responses, and 61% of people said subtitles. But I did not. You know, I say, like, don't listen to what people say, listen to what they do. So I looked at the analytics at YouTube and I realized that when a video is dubbed, people are about 40 to 50% more likely to watch the video towards the end, to the end than when it's subtitled. And so we were like, okay, then why are people saying they like subtitles? And so what we realize is these are just people that do not want to come off as, for example, maybe uneducated or having lack of knowledge of their original language. And they want to come off authentic. So it's really a virtual signaling Answer where they go. I want subtitles, especially in the world of anime, because they want to come off as sophisticated. And so this insight that I unlocked allowed me to even push harder toward dubs and say, no. Dubs is the path to move forward. Inherently speaking, we cannot watch a video and read text at the same time. It's a learned and acquired skill. If you're good at it, good for you. But most people aren't. It just puts a huge load on our cognitive abilities. Every single creator wants to be in, like, not every single one. Almost 95 plus percent of creators want to be in multiple languages because they know they are leaving money on the table. They don't know where to go. Companies that are in the transition space do not have the brand nor the insights that we do working with creators. So like I mentioned, creators just want to create more content. So the evolution we are seeing is there's a ton of companies right now that are working on AI dubbing, but they still force the creator to go and upload their own videos first. Choose your own video. So it's like if I'm a creator, I already have maybe hundreds of pieces of videos and there are 20 languages and then there are five platforms. So like, what videos do I localize for? What languages? And then I have to upload it on top of that too. So the title, the video description, all of that work. So really what we want to do is just make it easier and easier for creators to localize their videos and publish it, knowing that it is of good quality, if not great at like, you know, more and more affordable per minute rates. And so the evolution is a bunch of AI dubbing companies came. YouTube, Google Meta introduced a AI dubbing feature, which I know was vastly influenced by our work with MrBeast and other creators. And most of the demos that I'm seeing with AI dubbing today. It's like putting a car on a straight road with no obstacles, no pylons, no humans, no objects, and say, hey, it can drive itself. It's like, we watched this demo earlier today. This guy put like a video with no talking head in this system and said, oh yeah, it's AI Dub. It's like, yeah, dude, try to do an action movie where their background music and overlaps. It's not going to work, but it will work. It will happen. It just takes time. But what I'm not okay with is all these companies going to the market, rushing to the market and say, hey, you can paste the link and then out comes a dubbed video. In 15 seconds. But then again, it's just like, maybe be more communicative with your customers. Say, hey, this video that you're dubbing, AI dubbing, maybe one out of one out of every ten words are mistranslated. Or maybe the voice is not going to resonate with 20% of your customers. So at least give your customers the power to empower them to make informed decisions. And what we see with the creators and the transition companies, there is information asymmetry, where the transition company almost always knows more about the creator because the creator doesn't speak that language. So the creator depends on the translation company to tell them what videos, what languages, what platforms it needs to be posted on. And many times they're being taken advantage of. We. So we got the voice of Naruto for Mr. Beast Japanese. We got the voice of spider man for Mr. For Veritasium Spanish. And we also work with Mario Castaneda and Beto Castillo, who are the voice of Goku from Dragon Ball Z. So in fact, we did an interview with these two voice actors and it's so funny, it's actually on our unilingual YouTube page. You can go and check it out. We interviewed both. So there are two concepts. One is there are a very small group of people that think dubbing is not authentic. And it's like, okay, look, entertainment in itself is an illusion. Like, you think you go and watch a movie, you think that's authentic. Everything was made up by a storyteller anyways. The animation was made up by an animator to begin with. And so when it comes to dubbing, we are just doing the best we can to create the illusion that you're watching this video in your own language. So I get the expectations are high, especially with anime. And then there are a group of people that really want to hear the Japanese performance. But I'm like, hey, man, there is probably over a billion, if not a trillion hours of content on the Internet, ok? It's just a matter of do you want to consume it or not? If you don't want to consume it, if you don't want to have access to it, then you don't. And in terms of localized experiences in Spanish, so there are a few languages that do this. So Spanish, Arabic and Hindi are some of those languages where there are different regions speaking it. And because dubbing is like, the process is you first have to take the words of the characters, put it into speech or written text, then translate it, then feed those lines to a voice actor and the voice actor reads those lines over a video. And then you also have to do sound design. It's a lot of work. It's not easy. And. And so for Spanish, for example, Dragon Ball Z. I'm sure the title holder of Dragon Ball Z would have loved to create Argentinian accent and Colombian and then Spanish Mexico. But it's just the fact that it's just cost prohibitive and to create five different versions of the same dub financially cannot be justified. So you fall within. Again, so you fall within the people of that prefer authenticity. What I'm saying is most people, 9 out of 10 people fall outside this category where they just want to consume. They just want to be entertained. And Netflix ran a study where they took a French show that they dubbed for Americans. It was dubbed and subtitled. Every person most people they spoke with, they said they want to. They love the French accent and they want to hear the subtitles. Then YouTube, then Netflix tracked their behavior and said, oh yeah, when it's. When they. When we show them, by default the dubbed version instead of the subtitle, People are much more likely to want to sit and watch the video longer. The expectations are any company that promises AI dubbing should be more communicative and transparent with the creators. What I do not appreciate is them like, saying, hey, just come and paste the link. And it's like, hey, I'm an optimist when it comes to technology. I want my parents and my family to have access to world's best content in their language, which is Farsi, at a fraction of cost of what is today. So I have to be by inherently, I have to cheer for the advancement of AI. I just think that we need to say, like, hey, 11 labs, if you want to come out, or even YouTube and say, hey, we have an AI dubbing product. Communicate this with your users. The level of quality. Because, look, if I get a car and the car can drive itself, as a user, the moment that car drives into a tree, I immediately see the impact. But when I. When I dub my video using AI into Spanish and I don't speak Spanish, I do not know the full ramifications of what I'm giving to my customers. So I just want more transparency and I just want to see a very clear metric. Tell your customer how many words are actually accurately transcribed and translated. Also communicate with them. What is the transfer of emotion in the voice from one language to another? What percentage are you capturing? Are you capturing 70% of the happiness, 10% of the happiness, or 90% of the happiness? And then allow People to make informed decisions. I think a lot of these voice acts. So I have spent like months and months in dubbing studios in Mexico and Argentina. We built our own dubbing studio in Mexico City. I mean, when I say I built it, I mean I put out whisper rooms. We had three whisper rooms. We hired directors. We had voice actors coming in. We did the full thing. What I see is a lot of the times the voice actor has just driven for three hours to come to this random show that they were not briefed on, and they have to show up and put themselves in the shoes of the original anime narrator or the original actor. It's really a tough job. I think what we do is we just bring a bunch of onions, we cut it up right before the session, and we have them smell it, get them crying, or we get them to think about, you know, a parent, a loved one that passed away. No, no. I. I think what we're doing now is we are trying to stand on the shoulder of giants. I think OpenAI is doing a phenomenal job with synthetic speech. It is able to laugh, sing. As for on the human side, we are annotating the transcription. So we do two things. I realized that a lot of the time, as I mentioned, voice actors show up to a recording and have no idea who their character is. So we give them a show guide. We say, hey, this is your character. These are the emotions, and this is what this character in this story goes through. They are a kid, then they grow up and then they lose a parent and then they have to rebuild their life and then they become a successful entrepreneur, for example. So now the voice actor is able to put themselves in their shoes. So we create these show guides. Also we annotate the transcription. So we allow the voice actors to adapt quickly on the spot. But then again, we are working with OpenAI to use their voices and then in post production, we pitch the voice up and down. But I always said, I'll say it again. The idea of dubbing is never perfect. It's never going to be perfect. I also bet that I can go to an animation to a Naruto show and I can probably find a mistake where the animator maybe did not draw a perfect line, but we were just not able to tell because we're so immersed in it. It's creative work. It's never going to be perfect. Video and content is vastly different than when you're Uber and you have your Uber app and everything is well defined and you're translating it versus creative content is changes from Character to character. So to think that it's going to be perfect is an idea that we should never, at least for now, accept. But I think, yeah, in a few years from now, we will be able to fully capture the essence in the original in the source language and then transfer it through AI to the target language. Yeah, I mean, right now we were going to go down this path of let's build a voice recording platform and let in as many voice actors as possible. Then what we realize is, no, we really just need a few thousand voice actors, the best voice actors in the world. And also, for example, in Mexico, the dubbing industry is like the mafia. You have to go through many rounds to get in. We met this kid, 18 year old, not a kid, like this younger guy who's a voice actor and he came in, dubbed a YouTube channel for us. And the channel now has 4.3 million subscribers and 1.4 billion views. And now because of that, he's able to get into the dubbing industry. But I think the dubbing industry is going to, as it looks, it's going to look vastly different in a short few years. Anybody who's in the dubbing industry, like, watch out. Things are moving fast and a lot of the voice actors who are saying, hey, I don't want my voice to be trained on AI model, well, guess what? OpenAI Mid Journey 11 Labs, Heygen Runway, all of them have already trained their models on your voice and there is not much you can do. It's basically like technology moves faster than lawsuits do. Like by the time the legislation is placed. Yeah, it's like your voice is being used for a TikTok. Get over it. It's hard, it's sad. I get it. It's like. But the only constant in life, as we all know, is change. And things are just going to change. So we have just have to adapt. I really draw meaning out of serving humans at a large scale. So I just love that we are generating so much good in the world. Maybe not so much with Mr. Beast. I actually looking back at it, I was like, if I could push a button and undub all Mr. Beast videos, I would likely do that. I'm not sure what kind of good he does bring to the world. It's like, you'll never watch a MrBeast video. And after that thing, yeah, I'm going to go out to the street and donate $10,000 to somebody you probably won't like. Nobody probably has. But I think where I'm today is I am in the office like legit, just every day in the office, seven days a week. And my head is at engineering, like hiring engineers, bringing more engineers in. And just think like, okay, how do we use AI to bring the cost down to a point where more and more creators can use this? And this idea of free dubbing is good, but I think it applies. There's a power law in youth in the world of content where 5, probably 5% of all content in the world generate 95% of all views and revenue. And that 5% changes, it doesn't stay the same. And it probably is going to grow bigger and bigger with AI tools because now you used to have a build a team of five people to build a YouTube channel. Now you can just be one person, use 11 Labs for the voice, use Midjourney for the content, use ChatGPT for the script, and then you have a YouTube channel, which means the barrier to entry is going to be less and less and more creators are going to be coming and generating more views. So we have three values in a company that language. Language is just a tool to communicate. It's mere language is merely a tool to communicate, nothing else. It's like no more, no less. Let's look at it that way. So if it's a tool, then this tool needs to constantly be played with and disrupted and taken apart and put back together. And number two, everybody contributes to the final product. So for example, when a transcriber transcribes a video, if you do not catch an error early on, it's going to make its way all the way to the final product. Right. And number three, we really obsess with giving people a good experience, a hyper localized experience where they can come and connect with the content in their own native language. So the culture is, I would say we demand, we're demanding. But for example, today we had a team member where the family member unfortunately passed away and we're saying, hey, take as much time off as you need, it's fine, I get it, I've been there, it's okay. Whenever you feel like, you know what, you can come back, you know, it's a close family member come back. But yeah, I think right now we all know that internally we have to disrupt ourselves because if we don't, then we are going to get disrupted from outside. I just, I thank you for having me. I really appreciate it. Thanks for giving me this opportunity. [00:26:27] Speaker A: And this was our conversation with Farpod Mansorian, CEO and founder of Unilingo. We hope that you enjoyed this conversation and that you found it as inspiring as with it. My name is Eddie Arrieta, CEO at Multilingual. Thanks for listening. And until the next time, goodbye.

Other Episodes

Episode 295

January 01, 2023 00:10:06
Episode Cover

Life Sciences 2022: A year in review | December 2022

In the last installment of The Lab for the year, columnist Mark Shriner takes a look at the past year and predicts some trends...

Listen

Episode 58

April 01, 2022 00:05:11
Episode Cover

Report recommends providing language support training to UK prisoners

The Bell Foundation, a UK-based charity organization specializing in language education, has published a report highlighting the need for better language access services in...

Listen

Episode 13

February 01, 2023 00:06:08
Episode Cover

Book Review | Galateo Per Traduttori: Regole e segreti per un rapporto di successo con le agenzie di traduzione | January 2023

Galateo Per Traduttori: Regole e segreti per un rapporto di successo con le agenzie di traduzione (Italian for A Translator’s Etiquette: Rules and Secrets...

Listen