Transcript: Arcads
[00:01.053] 👩🎤 Ate-A-Pi: So Romain, what is ArcAds?
[00:06.485] 👨🏭 Romain: Our cards is a platform that helps marketers create winning ads with AI actors.
[00:15.453] 👩🎤 Ate-A-Pi: Okay, so when we, so the background to this is that about one and a half months ago, we had a super viral tweet, which I think was a actress or was an arcade, where there was a young woman, she was speaking into the camera and she was talking about a, I think like a body odor product.
And it was this classic TikTok style product promo speech monologue. And I had posted, this is AI. And that's when that had like 4 million views or something. And so let's talk about that a little bit. ArcAds produces these kinds of ads, basically.
[01:14.965] 👨🏭 Romain: Yeah, exactly. So the ArcAds team, my co -founder and I, we come from the world of basically growing consumer products. And when you are growing this kind of product, one of the main acquisition channel that you have is paid marketing. And when you run paid marketing, your key differentiator is the creative. The technical aspect of like making an ad,
has been made much simpler by the platforms now on Facebook ads, TikTok ads. You just go to the platform, select a broad audience. Now you think it's like the US, something like that. And you don't do those little tweaks to, at least they don't do such a difference like they did in the past. The thing that will move the needle for you is the creative itself. And it can mean...
how good is your video and also how many videos do you test? Because sometimes with paid marketing, the best way to achieve very high performance is not to try to brainstorm the best idea for an ad, but more set to test a lot of very different combinations until you find the winning one. So we built this platform to help marketers.
do that much faster and indeed the video you are talking about was created by one of our clients using Alcads to promote her product. And yeah, it went completely viral on Twitter, TikTok and other social media platforms.
[03:00.797] 👩🎤 Ate-A-Pi: Amazing. So let's talk a little bit about this ad itself. So this young woman who was actually an actress, her name is Ariel, I believe.
[03:12.565] 👨🏭 Romain: Correct.
[03:13.437] 👩🎤 Ate-A-Pi: So how does she become an ad? What is the workflow from you saying, OK, Ariel, come into our studio. How does she become this ArcAd?
[03:27.349] 👨🏭 Romain: Yeah, very good question. So, OIL and other, like other UGC creators are people who are very good at creating content that will feel organic and native to the platform. And marketers have been hiring creators to make their ads, not necessarily because of their audience, like in traditional influencer marketing, but more for their capacity of creating.
a video ad that will be very native to the platform. That's exactly why we created Arcade, because it's a very manual and long process to create this kind of content, but you still need to do it because on TikTok and on YouTube shorts and in the era of short form content, you want your ads to look like something that is not an ad. And that's actually the TikTok model. Like don't make ads, make TikToks. That's their...
biggest recommendation to make the performing ads on TikTok. So yeah, we help the teams find creators on our platform and generate the script without the whole process of finding them manually, talking with them, sending invoices. So the process for us to create an avatar is basically to find those talented UGC creators and ask them to send us.
a video of themselves in the most natural background that they can have. And actually we don't have a studio and we don't want to go in that direction because the content that you watch on TikTok isn't recorded in a studio, right? So yeah, so we'll take their phone. We have a couple of recommendations, like make it like you are doing a FaceTime with a friend when you send us the video to do the training.
We want them to be as natural and energetic as possible. And then the video that they will send us will set the energy for the actual video that will then be generated by AI using this input video.
[05:36.893] 👩🎤 Ate-A-Pi: Okay, so how long or how many videos do they need to send you? Is it like one minute, is it half an hour, is it an hour? How many videos do they need to send you?
[05:49.013] 👨🏭 Romain: Yeah, our recommendation is a two minute video, one video.
[05:53.341] 👩🎤 Ate-A-Pi: One two minute video. So Ariel would take a two minute video of herself in the car talking about some product or something that she's already done. It's not necessarily the product for this particular client. It can be anything.
[06:08.629] 👨🏭 Romain: Yeah, exactly. They will send us a video talking about anything. And then the interesting thing with Aurel is the first time she sent us a video, it was not in the car, it was a video of herself in her house. And you know, performance marketers, they take performance very seriously and they will test everything on our platform. So they will test the scripts for...
promotion of the product, but they will also test the actors. And that's the beauty of our platform. You can write one script and select 10, 20 actors, 50 actors if you want, and generate 50 videos and then put them all in Facebook or TikTok ads and take the results. And when they were testing the videos, it was very common that this actual avatar was getting very good results. What we did is we went back to Ariel and asked her to send us other videos in other situations. So Ariel in the car.
are you walking outside, et cetera. So yeah, the creator will send us, but each video, each individual video will be used for the generation of the future videos. It's not like we blend multiple videos together to make one training, but one training is made with one video.
[07:23.005] 👩🎤 Ate-A-Pi: Okay, so one two minute video of Ariel in the car can produce many marketing videos of Ariel in the car talking about different products. But it's not gonna produce a marketing video of Ariel on the beach. For that to happen, Ariel needs to send you one of herself on the beach.
[07:35.765] 👨🏭 Romain: Exactly.
[07:47.805] 👩🎤 Ate-A-Pi: Right. So, and basically what happens is, you know, once you identify a influencer or, you know, potential, you know, Arcad, you know, avatar, creator, and you say, hey, you know, would you like to send us, you know, a two -minute video, and then we're going to use it for this kind of thing. They send you a two -minute video. You put it on the platform. You train a model at that.
[08:03.029] 👨🏭 Romain: to.
[08:17.309] 👩🎤 Ate-A-Pi: at that point in time, I guess.
[08:19.413] 👨🏭 Romain: Exactly.
[08:20.573] 👩🎤 Ate-A-Pi: And the model is trained on, it's kind of like, is it a Laura? Is it a Laura video Laura, something like that? What kind of model are you training?
[08:31.829] 👨🏭 Romain: Yeah, so this is kind of our secret sauce and the space is becoming more and more competitive. So we'll would rather not share too much about like how the technology works in the, in the, in the backend.
[08:42.237] 👩🎤 Ate-A-Pi: Sure. Sure. And the voice, is it also, you're also training another model for the voice?
[08:50.869] 👨🏭 Romain: Exactly. So, yeah, the two main aspects of recreating an avatar will be being able to make a new voice that will match the initial tone and energy of the voice in the input video, and obviously replicate the face with AI.
[09:14.909] 👩🎤 Ate-A-Pi: Amazing, amazing. And after that, you have this avatar, you have a trained model, markers can come in, they can see 30, 40, 50 avatars, and then they can put in a script, I guess. They can say, like, this is my script that I want the avatar to read, and then they can pick, like, 10, 20 avatars, and they click a button and it generates, correct?
[09:41.653] 👨🏭 Romain: Exactly, the way it works is you would write your script, pick the actors, click generate, and in a couple of seconds, minutes, you will get the same script with all the actors that you selected. And then you will probably duplicate the script and generate the same video, but maybe with a different CTA at the end of the video or a different hook at the beginning of the video with the same actors. And...
[09:54.141] 👩🎤 Ate-A-Pi: I see.
[10:05.757] 👩🎤 Ate-A-Pi: Mm -hmm. Mm -hmm. Mm -hmm.
[10:09.269] 👨🏭 Romain: By doing this, you will be able to A -B test every single aspect of your script actors until you find the right concept that will get the best performance for your ads.
[10:12.253] 👩🎤 Ate-A-Pi: Mm -hmm.
[10:20.893] 👩🎤 Ate-A-Pi: Amazing. So, and at that point, the marketer, the marketer, I guess, would, you know, take the videos, plug them into their TikTok advertising channels, and then like run 10 or 20 of them, you know, with small budgets, see the, see the A -B testing performance, and then like refine, refine, refine, and then finally settle on something, and then like push that out on a bigger campaign, something like that.
[10:46.933] 👨🏭 Romain: Exactly. When you are doing performance marketing, the whole challenge is what is the concept that will overperform the other concepts that we have. So we developed this whole methodology to have our clients get results with AI, but it's basically about A -B testing scripts, A -B testing actors, and then even A -B testing the editing because usually what clients will do is they will generate the actor talking.
And then they will download the video and go into an editing app, maybe remove the background and put the actor on top of the product, maybe add some captions. And this also has a tremendous impact on the performance. So the same script with the same actor, but a different edit can perform very differently. So the best marketers out there, they understood that it's all about testing at scale to get results. And they like...
[11:22.653] 👩🎤 Ate-A-Pi: Mm -hmm.
[11:41.085] 👩🎤 Ate-A-Pi: Mm -hmm.
[11:43.157] 👨🏭 Romain: make sure to A -B test all of those variables in their TikTok ad accounts, Meta ad accounts, YouTube ad accounts to find the winning ads that will bring 90 % of the perf.
[11:55.965] 👩🎤 Ate-A-Pi: Amazing amazing. So let's let's let's talk a little bit about like the the the run -up that you guys had in I guess like in late March so I Late like March 27th 26th This thing goes viral. I'm guessing it went it went viral on Twitter first this this arcade So it went viral on Twitter and then what what was the run -up like? What were you seeing in the office at that time like?
all of a sudden you saw immediate increase in interest or what was going on at that time.
[12:32.309] 👨🏭 Romain: Yeah, of course. And by the way, I'm very curious because you actually tweeted at the spike, during the spike, what was, how did you discover Alcat and why do you think your tweets stood out at the time?
[12:50.245] 👩🎤 Ate-A-Pi: So I actually discovered it. I saw another creator called Linus Elkenstam, who is more on video and artwork creation. And he had discovered you from someone else. I think it was maybe your client directly who had tweeted it. And I think your client had tweeted just to promote the product.
And Linus had stumbled upon it and he'd done a little bit of research and he saw that it was AI generated and he tweeted out. And I kind of specialize in making things viral on Twitter. And my specialty is specifically AI. And so I saw an opportunity there because the video was, Twitter is doing this thing where they're pushing video.
And you can basically, if you have the right commentary around the video, videos go viral very quickly on Twitter. So the key is the captioning. You need to caption the video. And it's about finding a product market fit between your audience and the caption. So in my case, my audience is an AI audience. So every single thing that I see,
[13:51.573] 👨🏭 Romain: What do you mean the right commentary?
[14:12.989] 👩🎤 Ate-A-Pi: I have a generator in the back of my mind, which is saying, how do I make an AI comment about something? So when I saw this and I recognized it was AI, I decided, OK, let's just, you know. And I was very surprised. I was actually very surprised that it was AI. We'll talk a little bit more about that. So I was like, is this really AI? And then I checked through. And then I saw the video. And I was like, yeah, it has the artifacts.
So it is AI. And then I clicked through the website and I saw what you guys were doing. And I said, okay, this is actually AI. And I said, okay, I'm going to say, you know, this is AI. And then I said, you know, there's a meme going around every time we recognize something which humans used to do very well, but now AI can do, people go like, it's so over. So I then added that commentary, it's so over, and I posted it.
And then I started to see the immediate response because other people are fairly shocked that it was AI also. And then my audience tends to be very receptive to particularly this kind of thing. And I have a good feel for Twitter at this point, so immediately it launched. Yeah, and so that was interesting to see the takeoff. I knew what I was doing. It's not like luck at this point. So...
It's.
[15:43.413] 👨🏭 Romain: So do you mean the it's so over thing is a good way to replicate virality for future like AI improvements?
[15:53.341] 👩🎤 Ate-A-Pi: No, it's basically you have to have a product, you have to have a content audience fit. It's like a product market fit. So basically, if you look at the timeline, which is Twitter's flow, and you look at what's happening on the timeline, there are some recurrent jokes that people make. So for example, like, it's so over, et cetera. So if no one's made that joke for the last couple of weeks, you can make it one more time.
[16:15.381] 👨🏭 Romain: Mmm.
[16:21.437] 👩🎤 Ate-A-Pi: But if someone made the joke already like this week, if you make it, it's just going to sound like you're copying and it just sounds stupid. So you have to have a very good feel for the timeline to replicate the memes. And then if you do, then if you have your timing right, you just strike at the right point and then it can go viral.
I think, and you know, I've spoken to a bunch of creators outside. I spoke to Quebble Cop, the YouTube guy, you know, a couple of weeks ago. I think every social media has this kind of like metric that, you know, how to make things viral. There's a different formula everywhere. And, you know, to some extent it is possible to see in every channel, like there's some kind of metric that you can, that things that will catch.
And that's just for Twitter. So Twitter is very live. So you have to see what's on the timeline at that point in time. And everything has to be fresh, because 24 hours, tweets will just die. Your distribution will die in 24 hours after a tweet. So yeah, a bunch of those things. And then if you put those tricks together, and then you fire off enough tweets, you'll get something that hits.
[17:38.101] 👨🏭 Romain: Okay. Super interesting. Super interesting because to answer your question, obviously it had a tremendous impact on our growth. And as a company, we need to understand like what are the triggers to replicate that in the future. Obviously, one of the triggers was also that for some people, it wasn't clear whether it was fully AI generated or whether only parts of the video are AI generated. That probably...
helped push the virality a lot too. It had a tremendous impact on our growth. A lot of companies are starting from the use case, from the technology and try to find a use case. And we've seen that quite a lot with crypto, right? You have this amazing technology, people are very curious about it. And they try to find a use case for it. We really started from...
[18:28.541] 👩🎤 Ate-A-Pi: Mm -hmm.
[18:37.429] 👨🏭 Romain: the use case. In the past, we used to own a mobile app studio and when you own mobile apps, at least 50 % of your efforts are about marketing and finding the right creative. So we understood the pain that you can feel when trying to grow a consumer product. This is true for mobile apps, but this is also true for e -commerce, for education products, and for lots of other niches.
[18:38.749] 👩🎤 Ate-A-Pi: Mm -hmm.
[19:07.989] 👨🏭 Romain: So we really felt this pain. And when we sold this company in like a little bit more than a year ago, we started thinking about those pain points that we had and we saw this new technology coming and we kind of like matched the opportunities together by starting from the pain point. And I think this is our biggest edge in the market to really understand performance marketing and how to create winning ad and...
how to use creative in a way that will help you grow your consumer product. And when you started posting about it and other people started posting about it, our clients started posting about it. Obviously, a ton of other people started relating to this very specific problem and started reaching out, trying the product, obviously. So yeah, the impact was...
was amazing and tremendous. And I think before you experienced the true virality with a product that actually has product market fit, because we had experienced virality in the past with other products, but they didn't have such a high product market fit. So you can get 4 million views and then it will have a very small impact on your business. But when you actually are solving a very strong problem for a very large number of companies.
[20:25.725] 👩🎤 Ate-A-Pi: Mm -hmm.
[20:35.349] 👨🏭 Romain: then the impact is massive.
[20:39.101] 👩🎤 Ate-A-Pi: So let's talk about, so when was the company founded? The company was founded, I think like in September, is that right?
[20:47.669] 👨🏭 Romain: Yeah, we actually started working on it before, much before September. And we onboarded the first customer early January of 2024.
[21:02.877] 👩🎤 Ate-A-Pi: Amazing. So early January, 2024, you onboard the first customer and then March, this thing hits. At that point in time, how many employees were you like, you know, what was your rough size at that point in time, like before it went viral?
[21:18.453] 👨🏭 Romain: We work with a lot of contractors, external contractors, but in -house we are a very small team. We are all remote and based in France. Some people in our team are based in Asia, all over the world. And we are like a seven people team right now, like in -house.
[21:26.205] 👩🎤 Ate-A-Pi: I see.
[21:35.613] 👩🎤 Ate-A-Pi: Right, right on, right on. And so what, like, when this viral moment hit in March, what was the immediate impact like? Because I saw, for example, like, the site was, you know, you were struggling to keep up with the volume, like there's just a huge number of people trying it out. So what was the impact on the site? Like immediately, like you had like, you know, 100 extra people try it out and then you had to like freeze, freeze or something like, what happened?
[22:02.325] 👨🏭 Romain: Yeah, we had hundreds, very quickly, thousands, tens of thousands of people trying it out. But we were still at the beta phase. We weren't ready for that because we are on a very tricky topic, which is the topic of deepfakes. And I know this is a very controversial topic. And we want to act as this third party that will ensure creators' safety.
So when we were starting the product, it was just people around us in our network trying the tool. We started with, as I told you, we started in mobile. So we had founders from mobile app companies, either single apps like company, like dating apps who need to do a lot of advertising, also mobile app studios, these kind of companies. You will have a lot of them in France, in Turkey, in the US. They own multiple apps.
And they specialized, they specialize in paid marketing to grow all of those apps. So we started with these kinds of companies, mainly in mobile, but very quickly, other types of types of company were hearing about us. And we started to get like this inherent product led growth, just for people because in the area of like paid advertising, lots of people watch what competitors do. You have ad library. So if you go on TikTok ad library, you will see every single ad.
your competitors are running. If you go on Facebook ad library, you will see the same ads they are running on meta. So there was this kind of inherent virality in the product because when someone generates a video, they will have it on the library. Also because just we are solving such a huge pain for the customer that they will talk about it with other like marketers.
So we started having people outside of the niche of mobile using the product. E -commerce obviously is a very big one. And we had like this small inherent morality, but it was all controlled and we were able to individually talk to every single customer. During the first three months, Micro Thunder and I, we talked to hundreds of customers and we were basically just doing that. My call, my day was just call, call, call, call, call. And I had like 10.
[24:17.501] 👩🎤 Ate-A-Pi: Nice.
Hmm.
[24:29.525] 👨🏭 Romain: At some point, I think I even have 17 calls per day, just talking with customers and understanding if we are solving the right problems for them. So we were able to control this and it was from day one, the biggest focus that we had to make sure that we ensure safety for the content that is created. But when you get this kind of virality, you need to make sure that you just cannot onboard manually every single customer.
[24:47.837] 👩🎤 Ate-A-Pi: Mm -hmm.
[24:58.901] 👨🏭 Romain: Right. And this is the product that we have is the kind of product where you get initial traction from the small businesses. And then you will have marketers in bigger teams hear about the product because a small business owner, like it was the case on Twitter, will tweet about it. And other marketers who follow them will see it and will try to want to try the product. And you like,
[25:20.861] 👩🎤 Ate-A-Pi: Hmm.
[25:28.869] 👨🏭 Romain: You need this base of small customers to grow your product, smaller companies, but you will talk to every single bigger customers, the ones from the biggest companies. And in our case, with the virality, we got a lot of small business owners and we couldn't talk to everyone. So we had to shut down the product to make sure that we ensure full safety for the creators.
[25:42.173] 👩🎤 Ate-A-Pi: Mm -hmm.
[25:58.607] 👩🎤 Ate-A-Pi: Mm -hmm.
[25:59.765] 👨🏭 Romain: For Fondue.
on the platform and also on the technical standpoint. So this was the first point, the most important one, but on the technical standpoint, we weren't ready to manage such a high volume. So for all of these reasons, the impact had been massive on a business perspective and on the short term, it hasn't been as massive as it could have been if it would have been like in six months when we were fully ready to embrace that.
[26:26.813] 👩🎤 Ate-A-Pi: Mm -hmm.
[26:30.333] 👩🎤 Ate-A-Pi: Mm -hmm. Mm -hmm.
[26:32.949] 👨🏭 Romain: But in the long term, like it validated the need from our customers and just accelerated everything. We just understood that we would reach the goals that we had for the next three years. We will reach them much, much faster than we anticipated. So yeah, that's the beauty of Twitter. When you go viral on TikTok, you will get a massive like consumer addiction, not addiction, adoption.
[26:38.461] 👩🎤 Ate-A-Pi: Mm -hmm.
[26:49.437] 👩🎤 Ate-A-Pi: Mm -hmm. Mm -hmm.
[26:59.101] 👩🎤 Ate-A-Pi: Mm -hmm.
[27:01.621] 👨🏭 Romain: But when you go viral on Twitter, you get a lot of people from the tech industry, like companies. And I think that's a very interesting topic, like that now with Twitter, you can build a massive company on top of creators like you, who will talk about your product and who are followed by other business owners, marketers who have much more purchasing power.
and the one we will find on TikTok, right?
[27:33.629] 👩🎤 Ate-A-Pi: Yeah, it's interesting. I think one of the things is that I want to check with you is have you spoken to Ariel since all of this happened? So Ariel went like super, like she's everywhere now, right? And I saw she posted a reaction video later on. So what was Ariel's reaction? Have you spoken to her since?
[27:57.813] 👨🏭 Romain: Yeah, my co -founder did. Because when you are such a small team and you get such a massive group, like it's not normal to handle so much volume as a seven people team. So we had to like separate the roles. He spent much more time with her, like getting her reaction. But what he told me was like, it was very positive for her actually. And she's a very positive person. So she wasn't stressed out by at least.
She didn't reflect it to us by the massive wave of virality that she had. But yeah, it was very positive. She got a full -time job out of it. She also had her fiber profile blow up, which is like a... Yeah.
[28:44.061] 👩🎤 Ate-A-Pi: She got a full -time job out of it. That's cool. That's cool. So yeah, that just shows you. That's amazing. So Ariel, it's interesting because I want to kind of hear the ending part of the story, right? Because when things go viral on Twitter, it's just that moment of virality you have maybe a few days. And then after that, you don't know what the ending point of the story is. You don't know what's the forever after.
So I wanted to check on what was the Ferrera after. So it's great that she got a permanent job out of it. Is Ariel American or European? I don't know. I didn't really catch.
[29:26.421] 👨🏭 Romain: Yeah, most of our clients advertise in the US, even though we are based in France. So yeah, most of the creators we work with are from the US.
[29:30.397] 👩🎤 Ate-A-Pi: I see, I see, I see. So cool.
So one of the things that I wanted to explore a little bit was you talked about creator safety. So what do you mean by creator safety? Why were you concerned when you had this huge increase about creator safety? What did that mean?
[29:53.237] 👨🏭 Romain: Yeah. So, I recently saw this video on Reddit that someone did an ad for product. I don't remember which product it was, but for product for, using the face of an influencer, like a TikTok creator, it was a bigger influencer. We usually, we typically don't walk with big influencers at the moment. We, we instead walk with smaller UGC creators who.
don't have a huge audience, but are very good at creating content. But this one was a big influence. She had millions of followers. And there was this one brand that used a similar technology that we built to make her face say something to promote a product that she didn't even know about. Like the creator didn't even know about this product. And this is inevitable in the age of
technology, like people will use technology in a good way and people will use technology in a bad way. And if you don't have a third party in the middle, like between the technology, the creator and the advertiser, you don't have a company that makes sure only content that would be approved by the creator is created with her likeness, then you have a risk of like people abusing it.
So that's what I mean by creator safety. That's one of our main missions at Arcades to make sure that creating ads with AI is made in a safe way for the creators. And what I mean by that is that only content that they would allow to advertise is advertised on the platform. And how we do that is by creating a very clear contract that will...
highlight the things that we would allow on the platform and that we won't, that we signed with the creator, but also having a very strong moderation system that will automatically catch the kind of content that will not be allowed on the platform.
[32:07.517] 👩🎤 Ate-A-Pi: Right. And it could be something like advertising for a medical condition, like, a cancer drug. The creator doesn't want to be associated with like, hey, I have cancer, or I have an STD or something. They don't want to be associated with that because if you're a young person and you have like, your face was used for a herpes or HIV prevention kind of campaign, that's not really something that you would have wanted. So I guess that's the key thing.
[32:35.093] 👨🏭 Romain: Thank you.
[32:36.221] 👩🎤 Ate-A-Pi: Right on, right on. So after you had, as you went through, it's been kind of like a couple of months. In the past couple of months, what has been the process been like? Has it all been like just onboarding clients, increasing server capacity, onboarding clients, increasing server capacity? Like what have you been up to since that moment of virality?
[33:03.893] 👨🏭 Romain: Yeah, again, we think our biggest edge is that we understand very well the use case. So most of our time, this is the case for the whole team, has been spent understanding the key values that we are providing to our customers. And the mission of the company is quite simple. We want...
help marketing teams make winning ads with AI. And I think this is like a very interesting intersection between creativity and technology because creating an ad used to be very time consuming and require a lot of like equipment. And if you think about like the first ads that were created or the not maybe not the first ad, but the first video ads.
created on TV, for example, you would spend hundreds of thousands of weeks or months producing the ads and then started evolving towards using tools like on your computer to create ads much faster. And I think what we are building and adding AI in this workflow is the next step, the next natural step into ad creation. So...
Most of our time is spent on understanding how we can leverage future technologies, AI, to help marketing teams create winning ads. So yeah, to answer your question, it's been about like talking to our customers, obviously, improving the product and building the best possible product. But this always comes from the use case, not from the technology and from the needs of the...
[34:55.429] 👩🎤 Ate-A-Pi: Mm -hmm.
[35:00.661] 👨🏭 Romain: of the users.
[35:02.077] 👩🎤 Ate-A-Pi: Right on. What would you say are like the big technical things that you are looking forward to? Like when you see on the horizon, is it this kind of like, you know, chat GPT 404 or, you know, what are the big technical things that are you think are coming up that are going to be interesting for your sector?
[35:25.909] 👨🏭 Romain: Yeah, one very important one is the text to speech. So right now, and you probably know this topic very well, but right now the text to speech is at a very, very high quality level. It's really impressive. And sometimes you almost can't tell it's AI and the 11 Labs, for example, is a company that is making tremendous product.
Progress in this area. We saw the recent like to GPT for release And the voice was amazing. That was the thing that I was the most excited about when I saw the the video not the UX thing but more the quality and the emotion in the voice and our customers one of the things they have been requesting the most is how can we make the the human actor more engaging?
And you have multiple ways to do that. And one of the biggest way you can do that is to have a more engaging voice. That's one thing I'm very excited about that. How can you put more emotions in the voice? We have actually already found some techniques to optimize that. Like tweaking the, we like, we use some techniques to make it easier for the user in simple ways to make it more engaging.
so that's one thing I'm very excited about. Another thing I'm very excited about that, would be the next big innovation for us would be the possibility to integrate products in the ads. Right now we focus on like avatars talking straight to the camera and making them as engaging as possible. And I think just by doing that for the next, like at least 12 months or two years, we will have so many marketing teams grow much faster.
[37:12.701] 👩🎤 Ate-A-Pi: Mm -hmm.
[37:23.893] 👨🏭 Romain: And with editing, then they will integrate a product like removing the background and things like that. But on the longer term, having the ability to have the avatar wear your product or hold your product and talk about it in real time, maybe with like 3D objects that you would integrate in the video, would be a very, very important milestone to make huge progress. So yeah, overall.
making a first step will be making the human actors more engaging using the voice, using the movement, making sure the movements match the script, making sure the emotion, the way the talk is like very captivating. And then the second way, like in second way, like integrate the product itself.
[37:58.493] 👩🎤 Ate-A-Pi: Mm -hmm.
[38:05.373] 👩🎤 Ate-A-Pi: Mm -hmm.
[38:16.665] 👩🎤 Ate-A-Pi: So kind of like, you know, if they're talking about a pair of shoes to be able to pick up the pair of shoes or something, something like that.
[38:25.173] 👨🏭 Romain: Yeah, it's okay.
[38:27.973] 👩🎤 Ate-A-Pi: So when you speak now, are you speaking to some of the larger teams now as the business has grown? So now you have interest from bigger teams who are interested in the product now?
[38:47.605] 👨🏭 Romain: Yeah, that's the very interesting thing about this new wave of AI softwares. Again, companies like 11labs, small TikTok creators can use them. Small store owners can use the product. And also like big Hollywood studios can use it to like translate or dub movies. Or the New York Times can use it to like to read articles.
[38:53.949] 👩🎤 Ate-A-Pi: Mm -hmm.
[39:17.493] 👨🏭 Romain: And we got the exact same situation. We've got like very one person business, very small business owners who use the product to, for a different reason actually, they use it just to make their first 5, 10 ads, just for the sake of making the ads and having a way to make an ad in a cost efficient way and mainly in a fast way. But then when these users use the platform,
There will be the proof that it can work to generate engagement, which will trigger bigger companies to understand that now they can do much more content. And then the use case is very different. It's not about making your first 5, 10 ads, it's about scaling your ads. Because again, in the space of performance marketing, when you want to have very good performance, it's all about testing. And testing at scale with human is very hard. Basically everything that's involved.
human is very hard to scale. And yeah, that's what the platform allows.
[40:27.517] 👩🎤 Ate-A-Pi: Let's talk a little bit about founding in France. So you were also a co -founder of Class Growth, which was a company in New York. And so you had founded a company in the US as well. And before that, you'd been a founder in France. And now, again, you are in France again. What are the major differences between founding in France and founding in the US? What do you think the major differences are?
[40:58.453] 👨🏭 Romain: Yeah, I didn't spend much time in the US. But my opinion on this topic is that there is absolutely no difference. The company after ClassGrowth that we created was a weight loss app. And this app was targeting mainly the US. It was like the core market. 80 % of the users were in the US.
From day one, the app was in English. It was never localized in French. We never had a French mobile app. So, I mean, it's the same thing as like building stuff with a very small team. I think in 2024, you can build a massive company with a small team and also with a team that is all over the world. As soon as you understand where are your clients.
are and all of our fans, even the ones in France, most of them, they understood that the biggest markets for them are the US. So even then they will advertise most of the time in English, especially for like companies, mobile app companies. It's very uncommon that you find a mobile app company that is massive and that focuses on France. The interesting thing though is when you go to those big cities like New York,
or San Francisco or other hubs, it becomes much easier to find other people who are equally ambitious as you are. And when you are remote, this part is a bit harder because connecting over Zoom is not the same thing as having a small group of friends who are equally ambitious as you are. So that might be the difference that I can find. But I mean...
[42:39.261] 👩🎤 Ate-A-Pi: Mm -hmm.
[42:56.949] 👨🏭 Romain: If you go to Paris, there is an amazing community of entrepreneurs who are building amazing products, especially in the consumer space and especially mobile consumer actually. So you can find these kinds of things in France and wherever in the world. So far I didn't notice any major difference, but yeah, that's a very interesting topic. And I think the topic of where you are located.
as an entrepreneur, not as a company, but as an entrepreneur to meet other like -minded people is a very interesting topic.
[43:32.925] 👩🎤 Ate-A-Pi: Yeah, it's, it's, I think the, the, the thing that always surprised me is that, France has like some super mathematicians, right? Super mathematicians, super engineers, but then, but then the, the startup scene has been a little bit like, you know, behind in terms of like, even though the competency in terms of like engineering skill is like, is very, very high. The startup scene has like been a little bit behind, you know, not where you'd expect it to be.
And so is it a question of regulation? Is it a question of the market size? Is it a question of adoption? Is it a question, you know, so because the talent seems to be there, right? So.
[44:14.069] 👨🏭 Romain: Yeah, it feels like it's a question of market size because some of the entrepreneurs didn't start from the US as they should have. To me, the bigger companies, even in France, will be created starting targeting the US. And I have one in mind that has always been a very strong inspiration for us is the company you call Voodoo. They do mobile games. They are...
I think the third publisher in the app store, you have like the top trees like Meta, Google, and then you have Google. And they understood from day one that France wouldn't be a market for them. And from day one, they made their games for the US. And very quickly they were able to scale because they understood that. So I think that might be one reason. Naturally, obviously you will be biased to start with your local market when you start, but the...
the small proportion of people who understand that we'll get better results starting with the US. And another reason that might be something obviously is the funding. It might be harder to get massive funding outside of the US. If you want to get massive spending, at least I think the US is the best place to go. But in the world of like remote,
being in France isn't that of an issue to raise with American VCs. So I think it's all about the mindset. Eventually it's about understanding that the market is the US, the VCs are in the US, understand that as quickly as possible. But as you say, the talents are in France and that's the beauty of it. Like you can hire the best talent at a much more affordable price.
in France and still understand that the market and the VCs are in the US and build an amazing company this way.
[46:18.589] 👩🎤 Ate-A-Pi: Amazing. So let me ask you, so you're running a firm. Typically, what does your day look like? Because you have multiple time zones to manage. So are you up at 7 AM in the morning and taking calls at 2 AM in the morning? How does your day look like?
[46:40.917] 👨🏭 Romain: Yeah, so I will tell you the story of the day when we went viral. That was an interesting one. And then I will tell you about the normal day. So the day we went viral, it was around 3 a France time. So I woke up, it was around like 7 a I opened my computer and I see notifications everywhere. And I start understanding what happened. And then the day has been all about like managing this like...
surreal volume. That was the day when you go viral and you are not in the same time zone. That's what happens. But on a more day -to -day manner, most of my day will be about talking with the customers and finding ways to use AI to help them create winning ads. So it's...
about creating content to help them make winning ads, talking with them, sharing the value that you learn with other clients that might be applicable to them. So all about like helping our customers succeed and Micro Thunder, he's in charge of products. So he will be more thinking about how do you build the best product and the best experience and also the best technology to serve those clients. So all of those things you can do anywhere.
in the world, the calls obviously you will have to do them a bit late in the afternoon. Like we are finishing this call now in, and it's like around 7 PM. I have other calls today later in my time. It's just very long days, but when you are building something that is important, like I think we are, it's not an issue. Like at some point when we will like scale,
with enterprise companies, it might make sense to have someone in the US to be able to handle those conversations in person or to have more meetings and make sure that you manage all the customers' expectations. But yeah, at the moment, when you have a very good product, and I think it's all about making a good product, when you have a good product, you can do it from anywhere.
[49:09.245] 👩🎤 Ate-A-Pi: Right on. One question I have for you, what detail about your product or service do you actually care a lot about and you put a lot of effort into, but you feel like external users don't really get it or don't really pay that much attention to? You feel like it's really like, you put a lot of time into it, but it's not, you feel properly attributed that it's something that's special.
[49:36.597] 👨🏭 Romain: That's a very, very good question. Let me think about the main one. I have two in mind and I'm not sure which one is the most important. But I would say when we were creating this product, we also, there are two components of the product. The creation of the actual asset. When you do a video, you have assets and you have to edit those assets together. You have like, the assets can be...
The video, the A -roll and the B -roll. So the A -roll is the person talking, the B -roll are other videos that you put on top of it or that you mash up with the video. You have the song that you add on top of it. You have the captions. A lot of assets that will be at the end match together. And then you use an editing tool like Premiere Pro, CapCut or tools like that to edit them together. And...
Our initial idea has been that with AI, it will be much easier to create those assets, but it will also be possible to edit them in a much faster way.
[50:49.245] 👩🎤 Ate-A-Pi: Mm -hmm.
[50:50.229] 👨🏭 Romain: And our customers, they only care about, at least in our case, very good editing products that will be created probably in the next few years. But in our case, they really cared about the assets themselves. So how many actors do you have in the library? How many situations do you have like, Ariel in the car, but also Ariel working? They will care about what is the quality of the voice? What is the emotion in the voice? And making those as high quality as possible.
will be the key differentiator for us.
[51:24.733] 👩🎤 Ate-A-Pi: Indeed, indeed. Okay, Romain, so thank you. I think we're at the end of our time. So thank you so much. It's like a pleasure to catch up and to see you guys. I hope you guys do extremely well. Look forward to seeing more ads. I hope Ariel continues to work with you. I think the audience has an emotional connection to Ariel now.
[51:49.781] 👨🏭 Romain: Yeah, for sure. It was great. Nice meeting you.