I should know: I built your grandfather’s DAM.
We fielded our first Merlin DAM in 1993 (and that customer is still using Merlin). These days many industries are finding themselves drowning in digital visual and text objects and needing a DAM: the first industry to experience the digital asset overload was, back then, news organizations. They get as many as 60,000 images per day, and of course need to be able to rapidly search (using textual metadata, date city, caption, keyword, etc.) and put the best visuals to work quickly. DAM started in that high-pressure environment and has evolved ever since.
For the most part, it has not been that much of an evolution. About every 10 years new technology has shown up that lets us take a step forward: first of those were more capable search engines (since text was the only thing you could search against). Then storage prices fell and you could afford to keep a rich collection in your DAM. Video became commonplace and earned its place in your DAM. Lastly, integrations with other systems (like your CMS marketing systems, and social media) are essential.
But for the most part, your ability to find what you are looking for has not changed a lot in the last 20+ years: you have been totally dependant on textual metadata, like caption information, date shot, photographer’s name, etc. If many of your objects have sparse or no metadata, finding just the right visual object is really hard, especially if you have thousands or millions of photos, graphics, and the like. Paging through them can take forever!
That all started to change 2 years ago. The change was the flavor of AI called Deep Learning.
I’ve read other DAM companies writing that AI has been a disappointment: they refer to “disappointed and disheartened DAM customers” who concluded that “AI was not the holy grail to their asset tagging and discoverability woes.” That is certainly true of the vendors who rushed to use “off-the-shelf” general purpose AI licensed from one or the other of the huge cloud vendors, delivering useless tagging information from images (like “person”, “table”: how could that help you search for the right image?). instead of adding value they added noise, and charged you for it. No wonder those people were disappointed. It is kind of like going to your neighborhood pharmacy for delicate brain surgery. Good luck with that.
We took a different approach. We started with a deep understanding of how people use DAMs (some of us were photo editors before MerlinOne), we listened carefully to our user feedback, and then we read literally hundreds of academic papers to understand what AI is capable of, and what it is not capable of. We thought long and hard about what specific tools can add real value to someone working with images. We built a disciplined roadmap, a sequence of real, no kidding, time-saving tools that were never before possible. We selected a few approaches, found the right data scientists to work with, and spent the time to transform things that worked once in an academic setting into production-quality tools. There is a lot of work involved, extensive testing over huge collections of images (one customer has 8+ million images), performance, and accuracy tuning, but it has all been worth it. Each of these tools startled us, and we have been doing DAM for a long time, and we hope they will delight you too.
This blog is the first of a series. You cannot “binge-watch” them because I cannot write them fast enough. But if I do my job right, each week you will be entertained and educated enough about REAL, uses AI to impress your friends, and understand where DAM is headed.
This is not hype: we are betting our futures on it. It is an ongoing story, it will make your life measurably better, and parts of it approach magic.
Stay tuned.
Written by David Tenenbaum, CEO & Founder, MerlinOne Inc.