Moving Beyond Manual Tagging.

Semantic search is solving metadata tagging issues.

In 2019, while preparing for the D23 Expo, we adopted a different approach to our ingest and tagging strategy to avoid creating a media storage black hole where footage would be nearly impossible to locate later. The D23 Expo lasts three days and we anticipated shooting between 8 and 10 terabytes of video footage. Beyond ingesting that mountain of data, we were tasked with producing a daily event recap and potentially one or two two-minute marketing support pieces each evening. If you are accustomed to editing under very tight deadlines, especially during an event, you understand how important it is to be able to locate those moments that tell the story quickly and efficiently. That can be very difficult since you do not have the time to sit and review all of the incoming footage. We turned to Adobe Prelude to help solve that issue. Prelude was a decent product, but it was not the silver bullet we hoped for. Beyond the actual event, all of the footage we would end up shooting would be used for various marketing and publicity campaigns for the next event, which would take place two years later. Our hope was that Prelude would make finding those magic moments easier once the expo was in the rearview. That is not really how it turned out.

Adobe ended up discontinuing Prelude and shifted its pre-production workflow focus toward Premiere. Today, Premiere’s AI powered Media Intelligence addresses many of the issues we were trying to address with Prelude. Premiere now helps editors finding relevant shots without relying entirely on manual tags, memory, or long painful review sessions. This built-in AI search tool incorporates semantic search functions, which is what we were trying to achieve in 2019. At that time, we were limited by the tech and by a tagging process that depended on humans making judgment calls about what belonged in the metadata. Semantic search does not rely solely on tags, metadata, or file names. The system analyzes the visuals and returns results based on descriptive language, while still allowing transcripts, markers, and embedded metadata to support more traditional search queries. Unlike traditional keyword searches, which depend on the words available in tags, metadata, transcripts, markers, or file names, semantic search can understand more of what is actually inside the clip. It allows the editor to search for a ‘crowd reaction’, a ‘product detail’ or even a camera angle and actually find the desired footage, even if it was never tagged. Adobe Premiere analyzes footage locally, reducing the need to manually tag everything. This empowers the editors to search specific moments, scenes, or assets based on visual cues or spoken dialogue, removing the need to rely exclusively on metadata or file names. This gives editors and marketers a much faster way to locate relevant content without needing to look at everything that was shot. This reduces the need for an editor or assistant editor to take on the additional task of media librarian, allowing them to remain focused on the story.

The idea of semantic search is part of a much bigger philosophy of connected workflows. The concept of a connected workflow is built around the idea of linking all the pieces of the media lifecycle. A well designed connected workflow connects planning, production, asset upload, review, approvals, version control, and final delivery into a more unified process. 

By connecting those steps, an administrator will be able to track the progress of any project, locate potential speed bumps, automate communications, and significantly speed up the process. A connected workflow pulls ideation, creation, and iteration into a manageable process that is easier and more cost effective to scale. This allows smaller organizations that may have good footage and good intentions, but a fragmented workflow, to clean up their process and compete with much larger competitors. 

Technology has leveled the playing field for smaller companies that need to be nimble when competing with larger companies that have more resources. These new technologies reduce the need for the complexity of enterprise-level systems, but they still need to maintain the same level of discipline. The technology will not replace the need for file structure, consistent naming conventions, and a defined taxonomy. But that technology does allow smaller companies to turn every shoot, event, interview, and campaign into a reusable content library.

When preparing for the D23 Expo, we understood that we had three days to capture everything that would be used to market the next event two years later. Sifting through hundreds of hours of media over the next two years proved difficult, even for a company with Disney’s resources. Today, technology has made that struggle significantly more manageable, and those same tools are within the reach of companies of almost any size. At TMS, we focus on helping teams navigate this transition, turning disconnected footage into scalable workflows that make it easier to see and act on those missed messaging opportunities.