AnyClip Leverages AWS to Power its AI Video Analytics Platform
Think of all the apps, news outlets, social media platforms and brand sites you visit daily. How many minutes of video do you end up watching?
By some estimates, the average person watches almost an hour and a half worth of online videos every day. Now ask yourself: how many of them did you actually want to watch?
AnyClip is a Tel Aviv-based startup trying to make the answer “all of them.” The company has spent years building a platform that can analyze videos frame by frame, in real time, and identify what’s in them. The goal is to help media companies and businesses make their entire video libraries accessible, searchable and personalized on their own domains, making viewers spending more time on sites and not elsewhere.
To do it, they’ve built an ingenious AI engine using a series of AWS services.
AnyClip’s platform can analyze videos based on millions of parameters, taking in thousands of data points per minute of footage. The software recognizes people, brands, and products among others in real time. Those data points are pinned to timestamps, feeding into a range of tools from which users can draw actionable information.
A business news site can get a better grip on its library of content, searching by any number of criteria, and understand exactly what their readers want, and when. Advertisers who need to make sure their campaigns aren’t appearing next to something explicit or negative can filter anything out, or in.
Another use case is helping companies reach their audiences in smarter ways. “When it comes to user privacy, the world has changed,” says Gil Becker, President and CEO of AnyClip. The conventional advertising ecosystem has long been based on knowing users, he says, to target specific individuals or segments of them.
But AnyClip allows companies to create effective video strategies by getting granular on the videos, rather than just the audience. “We claim that marketers can target audience without compromising user privacy— they need to know about the content itself, as an alternative or in addition.”
Built on AWS
So, how does it run?
CTO Vadim Balannik says the company is using AWS to train their models, both for performance and automatic scalability. AnyClip stores its training data and models on Amazon S3, and uses the AI library MXNet in some of its models. Latency is a perennial challenge with video, and it’s been a focal point for AnyClip. “Service up time, this is of core importance to us,” says Becker. Hundreds of media companies, publishers, broadcasters and enterprises depend on their platform to keep their content running, and AnyClip depends on AWS as a reliable infrastructure provider.
Processing time is another factor, because the company needs the ability to scale up and down quickly. “Sometimes we need to scale out and to analyze thousands of videos concurrently, and immediately catalog, and sometimes time is less sensitive and we can scale down,” says Becker. The ability to set up new servers in minutes and then bring them down — and to do it economically — is crucial.
A new reality
Video may have already been expanding its online footprint for several years, but recent events have led to a revolution. “2020 has completely changed the way that we communicate—we communicate through videos, the events are only digital, and there are no on-premises anything,” says Becker. Compared to six months ago, all kinds of businesses are now having to manage and strategize over video.
He claims that companies’ objective is to make their audience come to their own sites to watch that content, rather than a third-party platform. “If you are a publisher, you want people to watch the videos on your own website, because this is how you make money from ads,” he says. “Retailers would want consumers to watch the videos on your own website, because this is how you sell tangible goods. Both are craving for the user data to retarget and get insightful analytics.”
For AnyClip, which had focused mainly on media and has recently expanded to businesses due to high demand, these changes are a seismic shift — making it even more important to keep building out its products on solid ground.