Animoto, with Boost from Amazon GPUs, Goes High-Definition
Animoto would have knocked my grandfather’s socks off. A freelance photographer, he spent countless evenings assembling his Ektachrome transparencies into multimedia slide shows—but back in the 1970s and 1980s, “multimedia” meant a pair of carousel slide projectors with a dissolve unit controlled by time codes embedded on a musical cassette tape. At Animoto’s site, by contrast, you can upload a few dozen digital photos and video clips to the Web, select some music, get back a professional animated video within minutes, then share it with your friends via e-mail, Facebook, or YouTube. It’s one of the slickest and easiest ways to package and share all those photos and clips from your last trip or party—and sharing, after all, is what photography is all about.
“It really is digital storytelling,” says Brad Jefferson, co-founder and CEO of the New York- and San Francisco-based startup, which is expanding into a gleaming new office on Kearny Street. “My co-founders come from the film and TV industry, and at the end of the day what we are trying to do is help people create short-form documentaries from the photos on their SD cards and computers. We like to call it ‘Hollywood production with the click of a button.’”
And just as in Hollywood, there’s a lot going on behind the scenes. It was around 2007, when Animoto was founded, that cloud-computing platforms like Amazon’s Elastic Compute Cloud (EC2) were beginning to make it possible for companies to offer such services online without having to build their own huge data centers. In fact, Animoto has always used EC2′s computing power for most of its hard-core rendering work—taking users’ photos, adding fancy motion graphics, dissolves, and other effects, and putting it all to music. But it was always a bit of a stretch, because Animoto was trying to get Amazon’s plain-vanilla CPUs to act like graphical processing units (GPUs), which are specially designed to speed the mathematical operations behind 2D and 3D rendering.
Now there’s an alternative: in November, Seattle-based Amazon Web Services announced that EC2 users could rent access to actual GPUs as well as CPUs. And that has allowed Animoto to take a big leap forward. Starting yesterday, Animoto upgraded its standard videos from 240p or pixels of height—the size of a small embeddable box on a blog page—to 360p, the size of a standard YouTube video.
Premium users previously had the option of upgrading their videos beyond that to 480p, which is DVD quality. But now they can also scale beyond that to 720p, or HD quality. And on top of all that, videos handled by Amazon’s GPUs are rendered about 10 times faster than before—and at lower cost.
“We knew that the direction our video rendering had to go was GPU, but the problem was that [until recently] there were no cloud providers with GPUs,” says Jefferson. “We looked at buying a bunch of servers and putting them in a colo [a colocation center]. But it was not having to invest in hardware that made us very nimble and allowed us to focus on what we’re good at, and we still don’t want to. Using Amazon GPU instances is the next evolution of that vision.”
There’s one caveat: Animoto offers 24 styles of videos, from abstract to cutesy, and so far HD videos and faster rendering are available for just one of them, the “Animoto Original” style. (For a look at that style, see the 2-minute Animoto video below, which I created last night using photos from a trip last weekend to California’s Sonoma County.) The reason is that Animoto has, in essence, …NEXT PAGE>
created a separate rendering engine for each style, and all of them were originally designed to run on Amazon’s CPUs.
Migrating the code so that it can run on GPUs is “really complicated,” Jefferson says, meaning it will take a few months to finish the job. But since about 25 percent of Animoto users choose the Original style for their videos, that was a good place to start.
The CPU-to-GPU transition has been in the works for nearly a year and has absorbed a good part of the 40-employee startup’s energy, Jefferson says. Embarking on the project was “one of those Innovator’s Dilemma sort of things,” he says. “But GPUs are what Pixar uses to render movies, and what gaming studios use to create first-person shooters, and if we are truly following on the heels of Hollywood we had to be on the platform that Hollywood uses.”
The upside to the switch isn’t just that Animoto’s videos look bigger and sharper. Now, the rendering time on a 30-second video, the service available to free “Lite” users at Animoto, is about 45 seconds, compared to 6 minutes before the GPU era. Six minutes is enough time to get up, grab a coffee, check your e-mail—and get completely sidetracked. The more-or-less instant rendering means Animoto users will be more likely to stay engaged and create finished videos that they want to share, Jefferson hopes. “Our expectation is that people will interact a lot more with their videos, because you are still in the mindset of what you wanted to do with it.”
Animoto’s roots are in the Seattle area, where Jefferson and his three co-founders Jason Hsiao, Stevie Clifton, and Tom Clifton all attended Bellevue High School; my colleague Greg profiled the company back in 2009, shortly after it raised $4.4 million in a funding round led by Seattle’s Madrona Venture Group. Other investors in that round included Amazon, Jeff Clavier’s SoftTech VC, and iStockphoto founder Bruce Livingstone. The startup has 2.5 million registered users and has been cash-flow positive since late 2008, according to Jefferson.
Animoto gained fame in startup circles in 2008, when it created a Facebook app for making free Animoto videos that quickly went viral, growing from 25,000 users to 750,000 in just four days—and sucking up so much rendering time that the company had to turn off some of the features that encouraged Facebook users to share the app. But since then, Animoto has de-emphasized its Facebook app, and it’s been more than a year since it updated its iPhone app.
That’s because neither platform proved very monetizable, Jefferson says: not enough free users converted to a premium plan. Instead, the company has focused on expanding its services to professional photographers and businesses. At the “Plus” membership level of $5 a month or $30 a year, users can make an unlimited number of videos, each up to 10 minutes long, and upgrade them to 480p for an extra $3 per video or to 720p for an extra $6. At the “Pro” level of $39 a month or $249 a year, the resolution upgrades are free, and users get access to a much larger selection of pre-licensed musical tracks. The Pro level of membership has proved extremely popular among wedding photographers, who typically resell the videos to their own clients, and realtors, hotel managers, and wineries, who use them for marketing, Jefferson says.
But Animoto still cares about average consumers, so it’s busy converting more styles to HD and creating better mobile apps—”This year we’re going to be investing very heavily in mobile, but making sure there are ways to monetize it,” Jefferson says. If Apple allows iOS developers to include an in-app subscription option in future apps, that could make Animoto’s iPhone app more profitable, as it would allow free users to upgrade to one of the premium plans, he says. An iPad app may also be in the works, but that depends in part on whether Apple’s anticipated iPad 2 includes both front and rear cameras, which would enable users to both shoot and assemble HD videos on the same large-screen device.
To handle all this work, Jefferson says the company plans to roughly double its staff this year, from 40 to 80, with about 30 engineers and marketers eventually working from San Francisco (as he always has) and the rest in New York.
Ultimately, Jefferson says Animoto is about helping users make snazzy, shareable videos without having to learn their way around complex desktop software like Apple’s iMovie or Microsoft’s Windows Movie Maker. Part of the startup’s advantage over potential rivals is a patent-pending “Cinematic Artificial Intelligence” engine that makes editing decisions based on the characteristics of a musical track—beat, tempo, energy, rhythm, and climactic elements—as well as attributes of the user’s images. “This whole process is part of our secret sauce to emulate what real Hollywood directors and editors do,” Jefferson says.
But while the visual effects may be handled by sophisticated algorithms, users themselves still have to capture the images, decide what order they should appear in, and choose a soundtrack. “It’s about creation, but creation as a means to an end,” says Jefferson. “What you’re really trying to do is share with your friends and family. When you sit down to iMovie or Windows Movie Maker, there is anxiety, because it’s work that has to be done. Animoto wants to relieve that creative anxiety altogether.” And provide a high-definition way to share precious memories.