AIMinds #023 | Quinn Favret, Cofounder, COO at Tavus Inc.

About this episode

Meet Quinn Favret, the Cofounder and COO of Tavus, a research company building foundational AI models that make generative video easy. They provide advanced text-to-video models via API for developers. Quinn's Michigan raised, and SF based, still doing his best to spend winters on ice skates

Listen to the episode on Spotify, Apple Podcast, Podcast addicts, Castbox. You can also watch this episode on YouTube.

In the all new episode of AI Minds, we explored Tavus, an AI research company transforming digital video creation. Quinn Favret, Tavus's co-founder, shared his journey from dropping out of the University of Michigan to founding Tavus and securing support from YC and Sequoia. Tavus leverages AI to create personalized digital twins and clones, simplifying video production. This innovation is revolutionizing sales, marketing, education, and entertainment.

Quinn's story begins in Michigan, where he enjoyed seasonal activities like swimming and ice skating. He started at the University of Michigan but soon found his passions lay elsewhere. In his sophomore year, he partnered with classmate Hassan, and together they explored various innovative projects. This led Quinn to leave university and join the TL Fellowship, further developing his entrepreneurial skills.

During this period, Quinn and Hassan co-founded Tavus, specializing in generative video through user-friendly APIs. Quinn's journey from Michigan's icy rinks to AI innovation marks him as a trailblazer dedicated to making advanced technology accessible globally.

Fun Fact: Quinn's vision for Tavus includes the ability for elementary school teachers to send personalized video readings to students at their individual learning levels, enhancing the educational experience with tailored support.

Show Notes:

00:00 Michigan native, dropped out, founded AI company.

05:28 TikTok offers personalized and relevant video experience.

07:52 AI videos integrated into various workflow tools.

13:10 Leveraging technology to expand customer interactions.

14:57 Customers prefer authentic communication over AI-generated content.

17:32 Demetrios, thanks for having us! Awesome.

More Quotes from Quinn Favret:

Transcript:

Demetrios:

Welcome to the AI Minds podcast. This is a podcast where we explore the companies of tomorrow. Building with AI top of mind. I'm your host, Demetrios, and this episode is brought to you by Deepgram, the number one speech to text and text to speech API on the Internet, trusted by the world's top conversational AI leaders, startups, and enterprises like Twilio, Spotify, NASA, and Citibank. We are joined in this episode by Quinn, the co founder of Tavus. How you doing today, man?

Quinn Favret:

Hey, Demetrios. Doing good. Thanks for having us.

Demetrios:

Well, I'm excited to jump right into a bit of your history, because I know you were a teal fellow. You dropped out of the University of Michigan to pursue what seems like a bit of a dream. Can you tell us how that happened?

Quinn Favret:

Yeah, it's quite the roller coaster of a story. So, born and raised in Michigan, so in the summers, I spent all my time in the water, in the winter, spent my time on ice skates and in trade, went to University of Michigan for college. And, well, I started there after a year or two. I realized that it wasn't necessarily for me, and at the time, my current co founder and I were building together and starting to work on just different random projects and things like that. So after my second year at Michigan, I ended up dropping out around that time, joined the TL Fellowship, and then that's when Hassan and I really started building Tavus as well. And it's, it's quite a cool story how we stumbled upon it. So maybe I can even take a step back and tell you a little bit about what Tavus is before I get into how we came up with the idea and how we came to where we are today. But Tavus is a research AI research company that focuses on generative video, and we really make this accessible to the world through APIs.

Quinn Favret:

So we allow developers and product teams to build human like experiences, video experiences, with really easy to use APIs. So what our platform really focuses on is creating personal digital twins or digital clones, allowing the end users to actually leverage their own likeness at scale. So, in simple words, you can clone yourself without ever actually saying a word or reporting a video, which is pretty dang cool. So, to be sure, maybe I'll stop there and happy to go in any direction you want, but wanted to give a little overview first before I dive too deep.

Demetrios:

What a journey. First of all, I want to hear all about the actual evolution of Tavus.

Quinn Favret:

In just a second.

Demetrios:

But I would like to touch on a few points. One is you started in the teal fellowship, you were creating and crafting the idea, then, then you went through YC, and then you went and got funding by sequoia. These are not like small names. This is some big names that are being thrown around. How has that felt all along the journey? And is there anything that you look back on and you go, okay, cool, we've done this. I imagine you're still probably thinking it's day one, but there's been some validation already.

Quinn Favret:

Yeah, the validation has been nice, right? It comes in waves, and I think it's the, the journey is certainly a roller coaster, and it has its fair share of ups and its fair share of downs. I think what we've been really lucky with, with all of our partners, whether it's YC or Sequoia or scale ventures, is we've just, they're exactly that. They're partners. And we've just found really, really great relationships with them where they've been able to support us and we've been able to work great with them. But I think what's ultimately really helped us is we focused on, one, our customers, what do they need, what are their challenges, and how can we solve for them? And then two, our product. We really focus on building the best AI models out there and creating the best experience around them. So it's those two things that have really guided us through it and, in my opinion, helped us to get some of this validation. But at the end of the day, what really matters is our customers using the product and how many of them are using the product.

Demetrios:

So talk to me a little bit about the inspiration behind the product itself and how you landed on realizing that there is a market need for people to create digital twins of themselves.

Quinn Favret:

It's a funny story. So we started building AI videos back in 2020 or so long before AI was cool. This is when AI videos were the big bad word of deepfake instead of generative AI, which feels like quite a while ago.

Demetrios:

That's so fun.

Quinn Favret:

We stumbled across the or, we came across the idea. Hassan and I were working on a different company at the time, and we were working with a few of our friends and companies trying to scale their companies, and we were helping them out with just different business growth things, whether it's email campaigns or others. And they were just having a really difficult time getting users, et cetera. So we suggested that they used videos. So one night, I remember I stayed up from 05:00 p.m. to 05:00 a.m. recorded 200 videos for one of them. Very similar.

Quinn Favret:

Just saying, like, you know, hey, Quinn saw you work at Tavus. Hey, Demetrios saw you work at Deepgram. Things like that sent him out, and the next morning got a ton of responses. And that was our initial validation, right. That video is really just a better way to communicate, a more effective way to communicate. Now, that was back in 2020. As the years have evolved, we've seen video become more and more predominant in today's society, especially short form video. You look at TikTok, Instagram reels, YouTube shorts, right? Becoming the main form of communication everywhere, which is quite amazing.

Quinn Favret:

And then what you're also seeing is a really high degree of personalization and high degree of relevancy. And that's why TikTok does really well, right? When you watch a video, it feels like it's for you, it's made for you, it's relevant to you. And what our product does in today's world is it bridges those two items. So we allow users to create hyper personalized, hyper contextual videos without having the typical pain points of actually recording a video. Where you have to get ready, you have to make sure your microphone is set up, you have to sit down and make sure that recording environment is good. Instead, you're really able to just sit down, and just as you would send an email, write a video script, and be good to go. Wow.

Demetrios:

Okay. It's funny that you mentioned that, because just before we started recording this, I was watching TikTok and going down a little bit of the rabbit hole and having fun with that. And that's how I like to take breaks sometimes and enjoy those funny, random videos and getting the remixes and all that. But it sounds like you started with the initial intention of doing this for salespeople. Is that a correct assumption?

Quinn Favret:

Yeah. So for the first year or two, we really focused in on marketing and sales experiences. So think about a workflow where if someone submits a form on a website, they would get a video back from maybe the CEO or the head of sales or someone like that. And the use cases there were really cool. We worked with mom and Pop, e commerce, Fortune 500 airlines across the board. It was quite awesome. But what we found is maybe in 2022, 2023 or so, that some of our best users and the ones we really loved working with were the ones actually tried to build with tavis. So they kept asking us for more API support and different features that would support them building Tavis into their product.

Quinn Favret:

So we made a decision in 2023 to really focus in on that audience and expose our APIs to the world where developers could start to build on our products. And that test was just amazing. So in today's world, we predominantly focus on empowering product and developer teams to build in AI videos, to their products, to their apps with our AI AI model. So it's been quite an exciting journey there as well.

Demetrios:

So how have you seen people using this? You mentioned building AI into their products. What are some of these use cases?

Quinn Favret:

There's some really cool ones. So we're seeing CRMs built it in as workflow tools, so they're distributing it to their sales and marketing teams using that CRM where they can just click a button within a workflow and it automatically adds video. We're seeing it with video communication platforms where they're adding AI avatar functionality. So instead of actually having to click record, you're actually able to just generate a video of yourself or create a scalable motion. And we're also seeing it quite across the board with video editors, different product teams, things like that. So it's been really cool to watch the adoption. And I think my favorite part about tabis is just watching how people continue to come up with ideas of where AI videos could be applied and used that we've never even thought of, or it's never crossed our minds.

Demetrios:

Yeah, I know. One is the education space, and that feels like a really big space that people can take advantage of, because if I'm trying to articulate something or I want to get some lectures out there, I can just create the lecture, and then I don't need to necessarily record myself and have to cut up the parts only that I like and all of that, I do see the ease at which I would hope this works. Right.

Quinn Favret:

That's awesome that you bring up the education space. One of the things we ask every employee who comes onto the Tavus team when they first join is, what's their vision for Tavus? And we found that everyone has a really unique vision. For example, my co founders vision is that when you get off the ride at, say, Disney World, there is Tony Stark Ironman saying, hey, Demetrios, thanks for writing. It was great to have you. Whereas my vision is that the elementary school teacher can send a personalized recording to each of their students reading a book at their learning level, which is such an engaging way of teaching. Right. So I love that you use that example, and I hope that in the next few years, that's a world we can live in.

Demetrios:

And so really, you're trying to just say, you know what? We don't want to throttle this product by deciding how people are going to use it and what verticals they're going to use it in. We're just going to create a product that anyone can call and incorporate into what they're building.

Quinn Favret:

Spot on. If we make it easy enough to build with, easy enough to use, our hope is that we unlock creativity with anyone who has ideas on how to use it.

Demetrios:

Let me see here. If I want to create a digital twin of myself, how much video and audio do you need of me before you can feel like you have a nice representation?

Quinn Favret:

Yeah. So it's a lot easier than you'd think. So we just need one to two minutes of audio and video and you sitting there talking. You can be saying anything. So you can talk about Deepgram, you can talk about Tavus, you can talk about demohops, anything you want. You would send us that video via API. Our system would take a few hours, process it, train it, and get you a really refined replica back, the clone of yourself back. And then from there, it only takes the API call to actually generate a video of it with using a specific script.

Demetrios:

Wow. And I can be wearing different clothes, or I can be shaved and unshaved with a mustache spa.

Quinn Favret:

Spot on. However. However you want the replica to look, that is all you.

Demetrios:

I do. So I can have my hawaiian version, I could have my Swiss Alps version.

Quinn Favret:

I've got my investor version, my customer version, my happy version, my friends version.

Demetrios:

Oh, that is so cool. Yeah, that makes it really nice. And then the background also. And where you're actually generating these videos, I guess I could be on the top of the Swiss Alps or the Matterhorn, or I could be in Hawaii.

Quinn Favret:

On the beach spot. Spot on. You could be, you could be walking around selfie style and have a TikTok style video. You could be at a really nice podcast recording studio and get something more professional, or you could even be in the office. So it's super, super dynamic, really. You can, you can put things anywhere. I think one of my, one of my favorites that we've seen is that we had a doctor on the work site who did a recording set up there, and I thought that was the coolest thing ever.

Demetrios:

Yeah, the creativity is unlimited. And you talk about creativity.

Quinn Favret:

I mean, it really is boundless. One of the examples I'd love to give is how people actually share and send the video. So when you think of sending a video normally, you probably think email, text message, embed it into a product experience. One of the coolest ones I've seen. We had a Fortune 500 company who wanted to reach out to an audience that they were working with. So around the holiday season, they actually generated 1000 videos and downloaded each one to a physical card like a snail mail mp4 player, and sent out the personalized video via mail. So it arrived in your mailbox, and when you opened it, it was a video from the person at that company saying, hey there, Quinn, it's been great to work with you this year. Da da da da da.

Quinn Favret:

And I mean, that puts my money where my mouth is here. When we talk about unleashing creativity.

Demetrios:

Exactly. The possibilities are really limitless, because it's just a matter of how I want to leverage this new technology that is at my fingertips and being able to, I can only fathom that when you're thinking at scale and you're thinking about things that you normally do with someone that you really like, or someone that is a customer that's very close or a very important customer, now, you can do that at a much larger level because you're not bound by, oh, I can only do this to the five people that I hold the most dear or that are my biggest customers, because you don't have to take that time. And so I could see it being for sure something that is on repeat. Are people aware that it's an AI generated video at or how does that, is it really real looking? So I may be fooled into thinking that. Oh, yeah. Like, they are actually reaching out to me. How does that go over?

Quinn Favret:

Yeah, that's a good question. With a complex response. So, on video quality, if I was to send you a video today of AI Quinn, you likely wouldn't be able to tell, but it was AI, which.

Demetrios:

Is you might be AI right now.

Quinn Favret:

I could be. I could be. We'll have a surprise for you in a few months here, where you'll find that pretty interesting. And it could be AI Quinn talking to you, but, wow. So, yes, you may not often be able to tell that it's AI, but where it gets interesting is the intent of the video. And the reason why the video is effective is not because it's fooling someone. It's not because it's tricking someone into you recording it. It's because it's providing well crafted content in a form that people prefer.

Quinn Favret:

So we have quite a few customers who actually disclose this was generated by AI, right. Or some similar message. And we've seen the results between the two style of videos to be exactly the same and identical, which, again, brings it back to that root hypothesis that people care about receiving video content where someone is speaking to them about something they care about. It's not necessarily about tricking them or making them seem like you recorded. In fact, if you do that, it could be a negative trait. Right. So we really want to empower that authentic communication here.

Demetrios:

So you are going back to the technical side of how you're building this. You're training the foundational models that are underneath the API.

Quinn Favret:

Yep. Spot out. So all models we do in house.

Demetrios:

Wow. And that seems like a very messy problem to get into.

Quinn Favret:

Theres a lot of new challenges that were working through as a team, even sometimes its the simple thing. What does it mean to have a high quality video, and how do you benchmark that? Quantitatively, theres endless challenges, but thats, to me, the most fun part. Weve got an amazing research team thats on board. Weve been very lucky and very privileged to work with some great minds out there on these foundational models.

Demetrios:

Well, Quinn, I'm so excited for everything you're doing. I am gonna get myself a digital twin right now and start playing around with it, see where I can outsource some of the different video creation that I do on a daily basis. And I'll get back to you with the results.

Quinn Favret:

I love it. Give it a sign up on the website, let me know what you think, and send me the digital clone version of yourself. There we go.

Demetrios:

So where do I go where I want to get rocking with this right now? What is it, Tavus AI?

Quinn Favret:

Check it out. It's Tavus tavus.io. And if you're interested in building with Tavus, just add a slash developer to that as well, and you'll have quite a bit of fun and be able to see our APIs. I always encourage people, check out the API docs. You'll see how simple it is and exactly how you can build with it and just dive right in and go for it. And if you do have any questions, feel free to reach out to me directly. I'm happy to help. Or jump on a call, too.

Demetrios:

Yeah, there goes my weekend, man. I'm going down the rabbit hole. This is great.

Quinn Favret:

Well, I'll tell you, generative AI video is a fun rabbit hole to go on. There's a lot of fun stuff going on right now, and especially when it comes to scaling that and deploying that, there's some really, really fascinating stuff going on.

Demetrios:

Excellent. All right, man, thank you for doing this.

Quinn Favret:

Awesome. Well, Demetrios, thanks for, thanks for having us.