Pond5, an up-and-coming media asset community and marketplace, employs proprietary algorithms to analyze images and footage for better tagging and search results.
For many editors, digital artists and filmmakers, having a ready supply of engaging stock visuals is a must. Of course, there's seemingly no end to the companies sprouting up in this $13 billion industry, offering very similar pools of content, even as the individual images and video clips differ from site to site.
As is often the case when competition heats up in the marketplace, these sites struggle to differentiate themselves from each other. Methods for drawing distinctions typically fall to marketing, with promotions for 'unlimited downloads' for a period of time and other special offers. I'm not knocking this limited time 'freemium' tactic or efforts like them, as they prove effective (otherwise, these companies would cease holding these promotions). However, when it comes to standing out, it's hard to beat innovation, especially those that make our working lives easier.
Pond5: AI-enabled tagging and search
A rising player in the media asset space, Pond5, has recently launched what it terms as "the media-asset world’s most effective artificially intelligent visual search engine." The video at the tail of this article outlines what exactly this means, but here's a brief run-down from company literature:
"Streamlining media discovery and search, this industry-leading technology utilizes next-generation algorithms to view and analyze videos, photos, and sounds. For artists, (product 'NextSense') automates the tedious task of adding searchable keywords to media files, by recommending a list of hyper-specialized keyword tags; for buyers, (product 'Visual Discovery') makes the process of finding the exact media file they need faster and easier than ever."
So, how well do these newfangled algorithms work in practice? Well, I jumped onto Pond5's site and played around a bit. From the welcome page, it wasn't immediately apparent that there's any tangible difference in Pond5's service. Upon making my first search ('puppy'), the search results seemed like what you'd expect from other sites. But, when I clicking on a particular image, the resulting suggested images results display rows of images with similar composition, color and content.
After clicking on several images, I was impressed how often Pond5's artificial intelligence got it right. There were a handful of results that didn't quite match, such as a field mouse popping up when you would expect a cat, but these instances were few and far between. For the most part, the search worked very well and I can see how Pond5 could be a useful tool for professionals that rely upon stock visuals.
I can't attest to how well the auto-tagging feature works, but the video below gives some clues as to its functionality. For those of you that rely upon or sell stock videos and images on similar sites, check out Pond5 and let us know what you think in the comments.