Where Do AI Tools Fit in Managing Media?

Earlier this week, MediaValet released a report on “Integrating Content Ecosystems with DAM and AI.” Before your eyes glaze over, this eminently readable report deals with a vitally important issue to media creators: How to make the most of our existing media?

We are drowning in media. Productions of all sizes are shooting hundreds of files consuming terabytes of storage. In the “Good Olde Days™,” I considered myself lucky if I could bring three cameras to a shoot. Today, it isn’t worth getting out of bed if your production isn’t using at least ten cameras, each recording iso. Sports uses cameras by the metric ton.

Yet, with all this fantastic coverage, the First Rule of Media still holds true: “If you can’t find it, you don’t have it.” There’s also the Second Rule of Media, in direct opposition to the First Rule: “Everyone hates logging!”

In their report, MediaValet makes the point that “Creative assets live in design tools. Campaign materials sit in project management systems. Approved files are stored in digital asset management platforms. Finished content is published through CMS platforms, email tools, portals, and social platforms. But despite this abundance, content rarely moves intelligently between systems. Why? Because integration frameworks are weak at best, and content ecosystems are not truly connected.”

They go on to argue that we need better tools for integrating these disparate systems. In fact, they define “content” as: “Asset + Metadata + Context + Rights + Expiry + Usage Rules,” all of which needs to smoothly and seamlessly travel from one system to another.

They outline the Five Layers of the Content Ecosystem:

1. Systems of Record. These support the foundation of all media: Digital Asset Managers (DAM), Product Information Management (PIM) and Content Management Systems (CMS).
2. Systems of Work. For example, project management platforms.
3. Systems of Creation. NLEs and DAWs.
4. Systems of Delivery. Content management systems like WordPress, social publishing tools like HootSuite and various distribution platforms.
5. Systems of Insight. These measure content performance and provide feedback on how assets are used.

I like much of their analysis; seamless integration is key for any workflow. But their analysis missed the mark – especially for content creators. Integration is important, but if the basement is leaking, you can’t build a very sturdy house. The Five Layers won’t work if the foundation of your media – the metadata that describes it – is missing or inaccurate.

Their report states AI tools are only needed to “extend the ecosystem by connecting additional tools, automating more workflows, and enabling AI-driven capabilities.” This  glosses over the essential first step of any asset management system: logging and applying accurate metadata.

AI can, I’m sure, help automate tasks associated with customization and distribution. But, remember the First Rule: “If you can’t find it, you don’t have it.” Shooting terabytes of stunning images is valueless if you can’t find the shot you need when you need it.

I’ve been a fan of Axle AI since I first purchased it in 2015. But, I immediately discovered that, while it was stunningly easy to add assets, it was hugely labor-intensive to label each shot and provide the minimum amount of metadata that would allow me to find it again.

In those early days, it was easier for me to remember the shots in my project than sit down and manually label them in a MAM.

Over the years, both technology and Axle evolved. Axle realized several years ago that AI-based tools can reduce the friction of applying metadata to assets. Harnessing the power and speed of machine learning to identify assets – actor’s faces, specific dialog, logos or objects in the shot – must be done before any automated system can be built.

Far too many productions have missing, inaccurate, or mismatched metadata which needs to be cleaned up for any DAM/MAM system to work effectively. I know from personal experience that manually applying and cleaning tags is never-ending, exhausting, and rarely accurate.

For example, in my library of 200,000 assets, I have multiple copies of the same asset, each with different metadata, image sizes, compression, and quality. Trying to manually find duplicates, standardize metadata and remove poor quality images is detailed, stressful and time-consuming. But, if that foundational metadata is not accurate, we can’t find the assets that we know we actually have.

Whether you run on Mac, Windows or Linux, Axle now provides AI tools that automatically add metadata to media via transcripts, face or logo recognition, identify elements in a scene, even apply custom settings to selected groups of clips. The old days of manually entering metadata quickly going away. Which means that once the foundation of your media – your clips – are properly stored and labeled, you can then build a powerful structure on top to help move media and all its associated metadata easily from one system to the next.