Asterpix is a company working on technology that adds so-called hotspots to video. These hotspots, which look like dash-bordered boxes, hover over particular objects in a video and can be triggered to show popup information when you place your cursor over them. The popup information has relevance to the object under consideration and consist of things like a description, related links, related videos, and theoretically advertisements.

Up until now, these hotspots had to be created manually by the content producers or publishers themselves. While they can still use Asterpix’s tools to do so, the company has begun automating the process by deploying bots that will find videos already posted on the web and using algorithms to tag them with relevant hotspots.

Asterpix bots are already crawling the various video sharing sites and hotspotting them at a rate of thousands per day. These indexed videos are being listed in Asterpix’s own video directory, which is provided through its site. The process is therefore mainly an exercise in testing and demonstrating the bots’ capabilities, since not a lot of people actually watch videos listed on the Asterpix homepage. The ultimate goal is to have video sharing sites like YouTube adopt the technology and index their videos with Asterpix hotspots automatically when users upload them.

So how are the bots managing to figure out not only the most important objects in videos but they popup information they are supposed to add for them? First, they judge the objects to hotspot depending on how long the camera focuses on them. The objects are essentially ranked by how much screen time they get. Then the bots determine the frequency of the terms used in any text associated with the videos. These are pulled from areas like titles and descriptions and are also ranked from most to least frequently used. At last, the bot matches the most frequent terms up with the most frequently viewed objects under the assumption that the two will match up appropriately.

Obviously this automated technique can’t provide the level of accuracy or relevancy that could be achieved by human input, but Asterpix representatives say that the system has been remarkably good at matching terms with the right objects.

We’ve embedded a sample video indexed by Asterpix bots below. You can also browse all the bot’s videos to get a better sense of its efficacy.

For another example of how interactivity is being added to video, see my piece from yesterday on Innovid.