The Future of Video Ads: Artist Controlled and Seamless
Sci-fi paints a bleak portrayal of the future of advertising. It’s in your face and intrusive. But, what if it was subtle and creator controlled? What takes us to that better future? One answer is Uru.
Uru wants artists and creators to choose the brands they associate with. The way Uru plans to deliver that promise may even put pressure on reviled ad units like occlusive pop-ups and unskippable interstitials.
It’s a win/win for content producers and viewers, alike. So, where did this crazy idea come from? And, how do they pull it off? The short answer, is the future-y magic of machine learning.
The long answer is in the following interview with Uru’s founders, Bill Marino and Brunno Attore. I had the pleasure of chatting with Bill and Brunno, to find out more.
Uru’s founders. Brunno on the left and Bill on the right.
Edit 2016.09.08: If you prefer audio, we’ve also uploaded a recording of this interview to YouTube. For our fellow podcast fiends.
Bill and Brunno share the vision of Uru
Q: I’m Paul, here with Bill and Brunno. And you guys are Uru Video, right?
Bill: Yeah, exactly. Uru.
Q: Ok, and what is Uru Video? What do you do?
Bill: What we do is we use computer vision to automatically find surfaces inside a video that can artfully host advertisement. So, we take any video as input and we’ll find the wall behind you, or the countertop in front of you. Someday, maybe the T-shirt that you’re wearing. And then we instantly and artfully blend advertisements into those surfaces.
Q: I think a lot of people in Tech view advertising as evil, given the popularity of AdBlock and other services like that. Do you feel with Uru that you’ve joined the Guild of Calamitous Intent, or do you feel like you’re doing good in the space, or making it better?
Bill: Do you want to take this, Brunno?
Brunno: I think the reason people don’t like advertisement is because it’s very obtrusive. It interrupts your experience, right? So, that’s the reason why we’re creating Uru. Because, Uru is a new way to advertise that actually does not interrupt your experience.
It’s something that is inside your video. It is in the background. It doesn’t go in your face. It doesn’t force you to click and close it. It’s just something that sits there in the background. It’s unobtrusive, it’s immersive, and we think it’s a much better experience for the user.
So, we actually see ourselves as the good guys, trying to change that evil.
Brunno shows off some of the tech. An MTV logo seamlessly overlaid on classic Bob Ross video.
Bill: Can I throw something in there?
Bill: I think we’re also . . we’re both artists. We’re both musicians prior to studying computer science. And, I think we’re coming at this from really a creator empowerment perspective.
We want to actually create, and we think this can [actually] exist. A world where creators work with brands they like to reach fans who are maybe interested in those brands and interested in those creators. There can be this magical confluence that happens. I don’t think we’d be working on it if we didn’t think that we could get there.
I have brands that I think are cool. And, if they’re being introduced to me by artists that I think are cool, then it’s a completely different experience.
So, that’s what I think we’re hoping to achieve with this.
Q: That reminds me of Henry Rollins talking about being a sell-out. And, he was like, “Do you not want any of your artist friends who are making music to make any money ever?” No. There’s selling out, which is selling your morals, and then there’s aligning yourself with something that you think is really cool and earning a paycheck. Everybody should get paid. I totally agree with that. I’ll see if I can dig up a YouTube link for that, for the blog version of this.
Brunno: Everyone likes some brands. Like, everyone likes to wear brands that they love. The problem right now is that the way advertising works right now is it’s very obtrusive to the viewer.
So, that’s the thing we want to change.
The two founders hard at work, at a NYC local Clarifai event.
Q: Yeah, I think how you guys are kinda shifting that landscape is: right now, advertising is determined by the publisher. So, it’s YouTube, or if it’s text, it’s blog site that you’re hosted on. They basically just put a slot on the page up to the highest bidder and that’s who gets it. Whereas, your approach is very much in the hands of the creator. The person who makes the video gets to choose what kind of stuff they want to insert into that video seamlessly.
Bill: Certainly, we’re headed towards a time when we’ll also deliver different ads inside these videos, per impression. But, we plan to work closely with targeting companies so that we really are trying to serve fans something that is very likely to interest them, and also be working with the artist to make sure [they like] the brands that are appearing in their videos. Which is a very intimate space, right?
The space that Uru creates is maybe the most intimate video advertising space. So, we have to work closely with creators, too. That’s the grand vision.
Q: That’s something really cool about your approach. No marketer wants to send an ad to someone who doesn’t care about their products. The holy grail of marketing is selling a product to somebody who wants to buy it. If they can just tap into that, that’s all they would do.
The reason why marketing is kind of like a shotgun approach right now is because we don’t have any better solutions. Brunno, actually, when we talked last week, you mentioned something about the origin story, or the idea for Uru Video. Where did that come from?
Bill: You want me to do it? I like to tell the full story.
Bill: It’s epic. So, me and Brunno met at the computer science Master’s program at Cornell Tech, which is the new New York tech campus that Cornell’s building. Currently, in the Google building. They were nice enough to donate a couple floors to get all the programs up and running.
We had done a couple projects together, liked each other, and then it became time to do David Tisch’s class. David is maybe the most notable seed investor in New York City. He teaches a class at Cornell that’s called Startup Studio. What you do is you create a technology, but you create it with eyes towards potentially spinning it out. Only some fraction of folks will actually spin it out.
We started working together and I actually convinced Brunno to work on a legal-tech (court data, actually) application. I was a lawyer prior to studying computer science. So, I sold him on doing that. And I think our goals were probably to do an amazing job in the class, get an A, work with really challenging technology and probably get jobs after graduating.
So, I sold him on working on this law-tech stuff, court data stuff. We were working on that. And within a month, we were just bored out of our minds.
Q: Really? With legal? I’m joking.
Bill: Yes, believe it or not. I know, it’s kind of sad. I hope there’s still hope for law-tech.
But, we were bored. Tisch was bored. The rest of the class was bored. We would present on it and it was crickets. So, we were like, let’s just find something that’s cool that we’re excited to work on. So, we’re just kicking around tons of different ideas, all kinds of ideas.
I had just read Ready Player One. Did you read that?
Q: I haven’t. But actually, some of the people listening may not have read it either, so give a little background.
Bill: It’s a dystopian novel about a not too distant future where we are in VR for large chunks of our day, if not our whole day. I believe, I’ve heard rumors that Spielberg is directing the film version in 2018, or something. They say that about every big book though.
So, I said look, when we get to a point, which may not be too far away, where we’re in VR or AR for large parts of our day, we have to think of new ways to advertise to people. It’s not going to go away. And, the present forms like television commercials, display ads . . they’re too obtrusive. There has to be more integrated ways of advertising. Integrated with the particular medium.
One of things that immediately pops into my head is finding these surfaces in the space and then swapping them out per impression. So, I’m inside a VR experience, I pass a billboard, it’s going to say something different than when Brunno or Paul passes it. And, you can apply the same thing to walls. Basically, any kind of surface. And, of course, product placement in the experience.
So, we were excited about that, but then we saw that if you just shift the gaze . . and, we haven’t dropped the ball on AR and VR, we continue to work on that. But, if you just shift your gaze a little bit and look at video, there’s a real reason to do this for video right now too, with these surfaces.
Number one, because it’s possible. Maybe for the first time, computer vision has reached this magical moment where what seemed impossible is now possible at scale. We knew that we could tackle this; finding surfaces inside video.
Also, I was a data science fellow at Mashable, online publisher, last year. And, it was when they started making a big push into video. I knew that there’s demand for new video ad units, especially ones that are less obtrusive. Just to circle back to what we were talking about, I see AdBlock as a function, or as a precipitant of obtrusiveness. If they weren’t obtrusive, maybe we wouldn’t be blocking them.
So, we saw the technologically feasibility, we saw the business reason to do it, and we just dove head-on into it. Is that an expression? Head-on? Head-long? I think it’s head-long.
So, started working on it like crazy. The way the class works is you work for a month, you present, and it snowballed. Everyone was really excited about it. We were producing amazing stuff. We filed our first provisional patent application on what we were working on in February. We wound up winning the “Most Likely To Succeed” vote in the class by a 30% margin. We won one of Cornell Tech’s startup awards, which is a $100,000 investment. So, things have been great.
And, that brings us to the present. I guess the subtext here is at some point, we could not deny that we had to work on this. The choir was too loud.
Brunno: I think we came to the realization that someone was going to do it, and why not us?
Q: Nice. Yeah, I can’t remember the name of the guy who said this quote, but, “Good sci-fi predicts the automobile. Great sci-fi predicts the traffic jam.” It sounds like that’s sort of what you’ve done.
It’s easy to think about the future in today’s context. Like, oh yeah, ads are blocking the content that you’re seeing. And, when you get into a VR world, it’s going to be right in your face. That’s kind of the Ready Player One universe.
What you did is take it a step further and say, “Actually, that totally doesn’t work in an AR or VR environment. It’s going to be much more subtle. It’s going to be projected onto surfaces there, as opposed to in your face.” That’s really cool.
How did you guys get involved in Cornell Tech? Were you learning machine learning there, or how did you start on this project?
Brunno: I can tell you. I’m originally from Brazil, so I really wanted to come to the US to get my Master’s. And in Brazil, I actually started my career working for fin-tech. I got a little, I can say bored with it. And, I wanted to get into the startup world. I joined a startup in Brazil. It was really nice. I really like the way things move faster and you are more dynamically [involved] together with the product and everything.
Then I started shifting my gaze towards working a startup, founding a startup. And Cornell Tech’s program is literally that. It’s like building your own products, learning how to run a startup, learning how to run your own company. So, it was a perfect fit for me. I came here, I was accepted. Came here, did my Master’s. Already had some experience doing machine learning. Took the classes there.
Cornell Tech has one of the best computer vision faculty in the whole country. You could say that really helped us in building our technology. We had awesome advisors. We had a lot of people that helped us into today. That really helped us get as far as we got.
Bill: There’s three Cornell Tech teams here at Friends of eBay. So, that tells you how fertile the environment is for creating viable startup ideas. Viable products.
Q: You went in thinking you were going to use this experience to launch some kind of tech career, right?
Brunno: For me, I always had a plan to launch a startup. I didn’t know it was going to happen this fast.
I think what Cornell Tech’s really . . it really excites about the startup world, because sometimes when you get there, you’re kind of afraid of, “Should I get a job or should I do this startup?” But then you start to learn things and see how it works. And, you get more excited. I think you can see a lot of people decide to run their own business, which is something amazing.
Q: Obviously New York’s been pretty hot on tech recently. The mayor’s office has a whole tech wing now. And, I’m fortunate enough to know you through Friends of eBay. I’m curious what made you choose Friends of eBay over all the other suitors you had courting you.
Bill: There’s a couple, I guess, serendipitous common interests. I know that eBay’s really interested in computer vision. They were excited to see what we were working on. And I know there’s a lot of computer vision stuff happening inside eBay. So that was a kind of serendipitous common interest.
But, it’s also a great brand too. We’re in the space of helping brands be even greater. I think that it would be cool someday to do something with the marketing or advertising folks at eBay. It just seemed like it was made to be.
Brunno: I think one of the things that I really like about eBay is that it’s something eBay’s really helping startups to do. It’s not something where eBay’s trying to take advantage of startups.
It’s like a good relationship where eBay’s helping the startups and at the same point, startups are showing eBay new ideas and stuff like that. That kind of environment is really helpful for startups, and that’s what brings startups to Friends of eBay.
Bill, sporting his Mastodon hoodie, is joined by computer graphics legend Don Greenberg.
Bill: Do you know, actually, also, eBay’s the very first technology I use every day. Do you know why?
Q: No, that’s interesting!
Bill: I don’t think you even know that. I wake up every morning and the first thing that’s in my Inbox is my alerts that I set on eBay. You may not know this, based on all my tattoos and so forth, but I am obsessed with J.Crew.
It’s true. Maybe we should keep this off the record.
Q: [laughing] I only ever see you in white t-shirts, man.
Bill: There are certain things that they make every year. What I do is I set up alerts, so when the rare times they appear on eBay, I want to be notified.
There’s other things too. There’s this Mastodon hoodie that I’m obsessed with. Blood Mountain hoodie. I have like 5 of them because I’m paranoid that they’re going to get holes and stuff.
So, I have alerts for those things. I wake up 5am every day, check my Inbox, see if someone’s posted the Mastodon hoodie. That’s how my day starts. Do you know this?
Q: That’s amazing. I’ve got to set up something like that. That’s a really good idea.
Cool. So, you guys were working on a legal tech application, and then you jumped into computer vision. Which to me, as a semi-technical person, seems like voodoo magic. Where did you start learning to do computer vision?
Brunno: Basically, Bill came to us and said [look at this book], and then we started debating it. This is a good idea, but we actually have no idea if that was possible or not. So, we started trying to play around with it, talking to people.
I started building started building prototypes and showed Bill. Everyday, “Look, Bill, this is what I’m getting.” Then, we started seeing that things were progressing. And I think- there’s a lot of tutorials on the internet for the basic stuff. But, as soon as we started to get more advanced, it started to get really hard for you to learn new things. As I said, we were lucky that we had a bunch of faculty members in Cornell faculty that helped us to get as far as we got. It’s not an easy journey. Definitely, you need to have some kind of machine learning experience. And definitely, you need people to talk to.
But it was fun. Overall, it was really nice to start this process of learning from the beginning, and like how to apply that. I think open source really helped us to do that. Like, using Open CV, which is one of the main computer vision open source libraries out there. That was really helpful for us.
Bill: Cornell Tech, no joke, although it’s like a nascent enterprise, they have one of the best computer vision faculties in the country already. It’s certainly one of their fortes.
Like Brunno says, when we started realizing that we were in the thicket, we just started pulling in professors. We literally met with like 12 computer vision professors or something, and just grilled them. That was where the leaps and bounds came from.
The cool thing about the Cornell Tech startup award is we get to continue to rely on that resource over the next year, which gives us kind of an advantage.
Brunno: It’s always nice, because actually we started with a . . (to Bill) should we talk about markers?
We started with a much simpler way to do our technology, which was setup to identifying services. There are things in computer vision called ‘markers.’
Which are basically images that are different from the rest of the frame, or the scenery.
Bill: It’s like a tag you put on something.
Above, Bill demonstrates how markers work. A standardized tag is placed on a surface, which computer vision algorithms then use to do their magic.
Brunno: And, they’re much easier for a computer to identify that than the scene.
So, we start with the idea of let’s do some kind of sticker that creators put on their walls, then we identify that and put ads on top of it, or something like that. Which seems like nice. It’s much easier to do than what we do right now. But that’s the big part [where] Cornell Tech also helped us. As soon as we started trying to talk to customers, talking to ad tech people, they said, “Look, no creator is going to put a sticker on their wall.”
We were like, “it’s over, we’re never going to be able to identify a surface.” The thing about startups is never giving up, right? So, we started trying to drill down on how to identify a surface. And we came up with a prototype which was very limited, and then we iterated on that to get where we are. We do a pretty good job of it, identifying surfaces.
Yeah, I used to work in the same building with Apiary. And, their CEO, I think his name is Jakub Nesetril said his philosophy on startups is it’s essentially a culture of amateurism. Where whatever you’re doing as a startup, you’re almost always going to be the first actor in the space.
You’re either creating a new market, or you’re up turning an existing market and putting a new spin on it. Which means you’re exploring a digital frontier that other people probably haven’t mined yet.
You’ve got to push forward with that. Do you agree with that sort of sentiment?
Bill: Yeah! Honestly, the weird thing is naiveté can be a huge advantage too.
Now, especially as we’re growing as an enterprise, we’re running into folks who have done ad-tech for decades, and they have a structured way of seeing the ad-tech landscape. I think us coming into it as novices and just having this hacker mentality is an opportunity. [It’s like] build towards it, if you hit a stumbling block, build around it.
I think that’s been a huge advantage, and I hope we can keep that. Because, this is a market that does really need disruption.
Brunno: It’s interesting, on that point. We talked to some veterans of ad tech. And, when they first saw our product, they kind of like, “hmm, that’s not going to work.”
Because, their standard doesn’t fit the way ad-tech or ad exchanges work. Blah, blah, blah, blah. And we’re like, “but that’s the old model. We want to create a new model.” I think part of being startups, there will always be people who will tell you, you cannot do this. You’ve just got to keep pushing forward and try to hack your way to what you want.
Q: Yeah, I like that point a lot, actually, Brunno. And I would even go so far as to say, as a startup, if everybody you tell your idea to says it’s a great idea, you’re probably not doing something daring enough.
That’s sort of the whole point, is to put your mark on the universe. Put a dent in the universe, as Steve Jobs used to say. You want to change things.
Brunno: And think, one more point, when we talked to those veterans in ad tech, some of them are resistant to the idea. But, the real thing that excites us is when we talk to creators, when we talk to people who actually watch the videos. They get really excited.
So, there’s something there, we just need to get there. It’s just trying to hack our way through the obstacles.
Q: Cool. So, people can go to uruvideo.com right now and see a demo of how it works, right?
Q: What are you working on now and do you have any dates in the near future that people should keep their eyes peeled for?
Bill: Definitely. We’ll probably do a pilot with top notch brands and top notch creators in September. Just put our technology in the hands of a couple folks.
Q: J.Crew, if you’re listening, there’s an opportunity here.
We think what we’re creating is magical. We have to kind of prove that, a little bit. We’ll do some heavy data tracking and show that . . and hopefully prove our gut feeling, which is that people respond better to this kind of advertising.
If these brands are there in a low-key kind of way during a positive experience of a video, then it pays dividends. And, that they’re more likely to visit websites and have positive connections with your brands. And, maybe even buy.
A really circumscribed pilot with the very best creators, the very best brands . . Get this thing out there into the world. That’s kind of what we’re aiming for in September. So, I hope come September or October, you’ll actually start seeing these videos out in the wild with our technology in them. I think for maybe January or so, or 2017, we’re going to aim towards something that’s a little bit more scalable. Hopefully a partnership with some video players.
There, what we’ll be doing is trying to use Uru to empower the users of all the different video platforms to monetize in a new way.
Bill and Brunno showcase the power of Uru at the NYC Computer Vision and Data Analysis Meetup. The UnderArmor logo is overlaid using Uru’s tech. Occlusion by the ping-pong ball is calculated applied for a seamless effect.
Q: Cool. And Brunno, we talked last week about leveling up the technology that you’re using on the back end as well. What new stuff are you pulling into your stack on that side?
Brunno: Last week, we basically got a neural net up and running. It’s really nice. We’re trying to basically create the whole infrastructure that’s going serve this trying to do these things in scale. Which is pretty good.
We’re setting up machines. It’s pretty impressive. Especially the cost of the machines. Because, we’re using very powerful machines for computer graphics and computer vision and they’re pretty expensive.
Q: Yeah, you’re actually using GPUs to do this, right?
Brunno: We have three GPU machines currently, right now. Each one costs [us].
Bill: But we have to shout out, IBM was really generous and let us into a program called IBM Global Entrepreneurs, and are heavily subsidizing that cost. So, thank you, IBM.
Brunno: Otherwise, we’d probably have run out of money already. But, it’s pretty cool. We’re using that. It’s really nice. It’s a really nice technological challenge for us.
Q: Cool. Well, that’s all the questions I had for you guys. Do you have any for me, or any shout outs you want to give?
Bill: Yeah. Thanks, Friends of eBay. We just got out of school. We’re raising a seed round which will stabilize us. But, having this program during this ramping up phase has been essential.
We’d be working out of coffee shops on our own or something. Being here in a supportive community has made a huge difference. So, I think that’s my main shout out. Thanks, awesome program.
Brunno: Same! Thanks, Friends of eBay. It’s really nice to be here. And thank you, Paul, for giving us this opportunity.
Q: Cheers, guys. Thanks for sharing.