What’s wrong with legacy methods?

I recently caught up with Maeve Campbell, a journalist working with international broadcasters. She uses multiple industry-leading media archives daily, and describes the current method as “digging for buried treasure with one hand tied behind your back.” 

The buried treasure chest is an apt analogy for legacy media libraries. The contents are unknown, retrieval is hard, and the value is locked away. The teams trying to utilize the archives are limited by the quality of their content discovery platform. Poor quality metadata and inadequate search tools put a low ceiling on your team’s capabilities. This means media archives aren’t living up to their revenue potential.

‍“Metadata, when managed properly, unlocks value that companies have previously left on the table” – Boston Consulting Group

‍Metadata is a critical bottleneck. Conventional search tools require every image, video, or audio file in a media archive to contain metadata describing its contents. This has required companies to hire people to manually label the data. This is slow, expensive, and prone to errors and omissions.

Once it’s created, static metadata risks becoming outdated. It can’t keep pace with real-world changes in language and search terms. For instance, “brat summer” was a cultural phenomenon of 2024 that spilled over from pop culture into politics. If your media was tagged in 2023 or earlier, searches for “brat summer” will overlook that potentially relevant content.

The limitations of static metadata mean that even after manual tagging, the vast majority of your media archive remains unlabeled or unuseful. The contents stay invisible to your search tools. Let’s explore an example…

Example: Shot selection like a Bat Out of Hell

You have multi-cam archive videos of a live concert. It’s a real belter. Meatloaf at his finest. But your manual metadata only specifies the artist, year, genre, and little else. How is your archivist supposed to find the moment Meatloaf walks on stage for the final encore, and takes a seat at the piano? Without super nuanced metadata, you’re relying on human archivists scrubbing through the video manually to find that precise moment. That relies on them knowing it exists in the first place. 

‍

Coactive's search tools help your team to pinpoint archive content with the speed and precision of AI. 

So far we’ve established that poor quality metadata is holding media archives teams back, and preventing media companies from monetizing their content libraries efficiently. But your teams are also suffering from crude search tools. In their eyes, that’s where the real friction kicks in…

‍

Search tools aren’t created equal

“Content is the reason search began in the first place.”—Lee Odden, Marketing Agency co-founder and author

The frontline users of your media archive, like archivists, analysts, and producers, can only work as well as your search tool allows. But conventional search tools are… how to put this… a bit crap?

They rely on keywords, which restricts the number of parameters that can coexist in a query. For instance, an archivist might search for an image or video with this string: “Empire State building at night”.

Try adding more variables and the whole thing face-plants. Zero results. So your team has to try shorter queries. Which means super broad results. Which means manually filtering and refining, not to mention learning all the different taxonomies. With their results whittled down from thousands to hundreds, they now do the final stage: manually scrolling and scrubbing through videos and images until they eyeball something relevant. 

Pass their desk and you’ll see the joy fading from their eyes…

Case study: how legacy search is failing journalists

Let’s take a real example. Maeve Campbell is a busy journalist (you met her earlier). She works to daily news deadlines. Every piece requires multiple images and videos relevant to that day’s news story. To find them, she has to search the company’s archives, as well as providers like Getty and Shutterstock.

“This morning I was prepping a story about climate change, and its impact on the islanders of Tuvalu. I needed photos and historical footage of the islanders’ daily lives, and specific camera angles showing rising waters affecting local buildings. The searches took forever - because the platforms delivered results like rogue holiday snaps, different islands, the wrong communities or time periods. Narrowing down the parameters and excluding irrelevant results was incredibly time consuming, but essential for the credibility and impact of the news piece. This is a daily pain for us journalists!” – Maeve Campbell, climate journalist

Traditional search tools force media archive users to do a ton of sifting for every query, which wastes time they could be using to add value elsewhere. Surely in the age of machine learning, there’s a better way?

‍

Introducing content discovery powered by multimodal AI

The sharper the search tool, the less time gets wasted digging for content. Content discovery with Coactive’s multimodal AI is like the precision switch from navigating by the stars to using Google Maps.

Coactive searches the vast uncharted landscape of your unstructured data, and retrieves highly-relevant assets in seconds. The Coactive platform is radically faster and more accurate than conventional methods.

We make this possible with three cutting-edge tools: semantic search, image-to-image lookups, and our groundbreaking Intelligent Search feature. Find highly specialized video, images, and audio assets from your archive with a few clicks. All accessible in one place, through our intuitive UI or via our API to your interface.

Let’s take a look at each feature, and draw some well-deserved flattering comparisons along the way. (For example, for fans of the Harry Potter franchise, this is like being able to say “Accio!” in real life.)

#1. Intelligent Search: Coactive learns your lingo in seconds

Every company has unique terminology and shorthand ways of expressing crucial ideas. Here are some of our favorites:

• Tech companies like to say “dogfooding”, when they need employees to test their own product. Yummy.
• Sprint likes to combine friends and family with the term “framily”. Ick.
• LinkedIn uses “ninjaing” to describe people solving tasks with high skill and efficiency.

OK this isn’t a Buzzfeed listicle so we’ll stop there, but you get the gist. A conventional search engine would return images of dog food, families, and ninjas, or nothing at all. 

Internal jargon has been a barrier for conventional searches, which can’t associate your phrasing with the relevant image, video, or audio in your archive.

With Coactive’s Intelligent Search feature, you can quickly train the platform to learn what you mean by a specific search term. No technical expertise required – just use a natural language query and label a handful of images or videos (typically 20 or less, often as little as three or four) that match your unique terminology. It only takes a couple of minutes to train a concept and get scalable results. 

‍

‍

Heads up: generic AI search tools like AWS and Google don’t allow you to do this kind of intuitive fine-tuning with no coding skills. Coactive AI gives your whole team the power of AI search in minutes. 

Example: advertising client needs merchandise footage

Say an advertising agency has a major client that prides itself on community outreach. They’ve asked your team to find social media footage of the client’s employees volunteering at different events. To find the right visual data, your search tool needs to understand what that company’s uniform looks like. 

With Coactive this is easy. Start by creating a "concept" (e.g., “Client3_uniforms”) and run a natural language search to identify images of employees wearing the uniform. Provide quick yes-no feedback to tag the images so the model understands your specific needs. Within minutes, the model is trained and allows you to efficiently search your entire dataset for images featuring your client’s distinctive uniform!

I think that merits a flattering analogy, don’t you? For fans of the sitcom Friends, this is The One Where You Get Superpowers At Work.

#2. Natural language search

Traditional search is like a public bus. It’s slow and you’ve still gotta walk the last fifteen minutes. Coactive’s natural language AI search is like a chauffeur service to your doorstep. Fast, clean, and no funky smells. 

In the example shown, the semantic search is for a “Basketball player in a yellow jersey hanging off the rim.” 

With a conventional search tool this would have required breaking the user’s request down into one or two keywords, like “basketball player” then adding filters like “people” and “contains color = yellow” and painstakingly scrolling and further refining to find relevant results. 

Coactive’s semantic search allows the user to search with natural language. Talk like you would to a human, and achieve highly relevant search results powered by AI - avoiding the need for detailed metadata and complex filters.

For fans of Rick & Morty, this is your very own Mr. Meseeks.

Example: TV producer needs a celebrity montage

In this scenario, the busy team at Let’s Talk Comedy has been tasked with creating a montage for tomorrow’s show. They have a Hollywood hero coming on the program, who is famous for their raucous laughter. The producers want a montage of the celebrity laughing, doing spit takes, and creasing up.

Your assistant editor opens the media archive and conducts a natural language search powered by Coactive AI. They type in: “Close-up videos of [celebrity name] appearing on talk shows and laughing or spitting out drinks next to other guests or hosts.”

Reader, they get it. A montage that would have taken hours to compile is done in a matter of minutes. Your expensive editor is now freed up to do their next task, and keep adding value. 

#3. Image-to-image search

‍

Ever heard a colleague say, “I’ll know it when I see it,” while struggling to describe what they’re looking for? This is their dream come true. It’s all show, no tell.

With Coactive AI you can use visual prompts to search your media library. This can help users to find close matches without having to consciously articulate all of the aspects they’re looking for in the results. 

Doubly cool: our reverse image lookup can be used as part of Coactive’s Intelligent Search feature, where you’ve taught Coactive the meaning of terminology unique to your business. Let’s check out an example.

For fans of Game of Thrones, this is House of the Dragon S1. Even better than the original.

Example: archive team need to replicate winning content

How fast can you find and replicate your winning content? Imagine an iconic image or video goes viral on social media. Your team is racing against competitors to find similar archive content and inject it into your campaigns or productions. With legacy tools, that search process is a slog.

With image-to-image search, your team can find more of what works, faster than ever. While your competitors continue wading through their archives, you’ve located the valuable assets and brought them to market.

Summary

Coactive's multimodal application platform offers a revolutionary solution for media and entertainment businesses seeking to monetize their archives. By leveraging advanced technologies like semantic search, image-to-image search, and Intelligent Search, Coactive helps companies to:

‍

• Discover your content’s full potential: _{bypass metadata bottlenecks and uncover incredible nuances in your content archives}
• Streamline workflows: _{equip your teams with more intuitive, accurate, and faster search capabilities}
• Leverage your own terminology: _{train our platform on your unique concepts and needs with just a few clicks}

‍

Coactive's platform empowers media companies to efficiently find, access, and monetize their valuable assets, giving them a competitive edge in today's rapidly evolving industry.

Ready to see our platform in action? Book a demo.