AI-Powered Search Turns Unstructured Media Libraries Into Queryable Data

Shade uses computer vision and custom AI models to automatically tag and organize media files, making hundreds of thousands of assets searchable.

Creative teams generate enormous volumes of unstructured data. Video files, raw images, graphics, documents—all sitting in folders with inconsistent naming conventions. Finding a specific asset means remembering where you saved it or scrolling through thumbnails for hours.

Shade, a New York startup with $5 million in funding, applies AI to make this data actually useful. The platform automatically tags assets, extracts metadata, and enables natural language search across entire media libraries.

While presenting to the 64th IT Press Tour, CEO Brandon Fan demonstrated how the system transforms chaotic file collections into organized, searchable repositories.

The Metadata Problem

Media production companies typically manage hundreds of thousands of assets. A single campaign might generate thousands of photos, video clips, and design files. Without proper organization, that content becomes a liability rather than an asset.

Traditional approaches rely on manual tagging. Someone has to watch each video, note the shot type, describe the content, and add keywords. This takes hours and rarely gets done consistently.

Fan explained the challenge: "The average media production company has hundreds of thousands of assets. Imagine trying to organize all of that manually. People spend more time searching for files than actually creating."

Shade's solution turns metadata generation into an automated process. Upload a batch of files, and the system analyzes them using computer vision models. It identifies objects, people, settings, and actions. Then it writes that information into structured fields you can query.

How the AI Works

The platform uses several AI models working together. OpenAI's vision models handle the initial analysis and description. For images, Shade can identify shot types (wide, medium, close-up), lighting conditions (day, dusk, night), and visible elements (logos, products, locations).

For video content, the system generates transcripts using Assembly AI. This makes spoken dialogue searchable. If a CEO mentions quarterly results in a video from two years ago, you can find that exact moment.

Facial recognition runs across all visual content. This helps teams track which people appear in which assets, which is useful for managing talent contracts and usage rights.

The search engine itself is custom-built. Fan noted that standard models couldn't handle the scale: "We took open-source vision models and distilled them to run efficiently on millions of assets. We needed something that works on CPU infrastructure to keep costs manageable."

The result supports semantic search. Instead of keyword matching, it understands intent. Search for "people playing cards in a park" and it finds images matching that scene, even if those exact words never appear in any filename or tag.

Organizing at Scale

Beyond search, Shade offers a spreadsheet-like view of entire asset libraries. Every file becomes a row. Columns show AI-generated metadata alongside custom fields you define.

This turns unstructured media into structured data you can sort, filter, and analyze. Which photos were shot at night? How many assets feature your logo? What's the total storage used by each project?

The platform lets you create custom metadata prompts. Tell the AI what information matters for your workflow, and it extracts that data from every relevant file. A consumer brand might prompt: "Identify visible logos and products." A real estate firm might ask: "Note architectural style and condition."

These automated tagging systems learn from your patterns. The more specific your prompts, the more useful the metadata becomes.

Business Intelligence Applications

The organizational capabilities extend beyond finding individual files. Teams can analyze their content libraries to answer business questions.

How much content do we have for each product line? Which campaigns generated the most assets? What's our storage cost trend over time? Where are the gaps in our visual library?

Fan demonstrated this with a customer example: "We have construction and architecture customers using this. They shoot property videos, and the AI tags conditions, styles, and features. When they need comparable properties for appraisals, they can query the system instead of searching manually."

The same approach works across industries. Healthcare organizations tag medical images. Retailers organize product photography. Security teams categorize footage from multiple camera angles.

Integration and Workflow

Shade includes webhook-based automation similar to Zapier. When new assets arrive, you can trigger workflows that route files based on metadata, notify team members, or update external systems.

The platform exposes an API that makes all metadata queryable. Development teams can build custom dashboards showing asset metrics, generate reports on content usage, or integrate with other analytics tools.

For organizations already using tools like Salesforce or HubSpot, Shade becomes the file layer. Instead of attachments scattered across email and cloud drives, all media lives in one searchable repository with proper tracking.

Cost and Scale

Shade charges for seats and storage, with pricing typically 30% lower than the combined cost of separate tools for storage, review, and search. The average customer uses 10 seats with 25TB of storage for about $10,000 to $15,000 annually.

Customers can bring their own S3-compatible storage from providers like Wasabi, AWS, or Google Cloud. Shade layers its metadata service and AI capabilities on top. This approach keeps data under customer control while adding intelligence.

The platform currently serves 94 customers, with most growth coming through inbound traffic and referrals. The company projects $10 million in revenue next year as it expands from creative teams into broader corporate marketing and document workflows.

Looking Forward

The next phase focuses on document intelligence. While video and image analysis work well, organizations also need to extract insights from PDFs, presentations, and spreadsheets.

Shade also plans sub-clipping functionality that searches within individual videos. Instead of finding the full clip, you'd locate the specific 30-second segment where something happens.

An upcoming app marketplace will let third parties build analytics tools on top of Shade's data layer. Think of it like Salesforce's AppExchange, but for file intelligence.

What This Means

Organizations generate more unstructured data every year. Without proper organization and search, that data has limited value. You can't analyze what you can't find.

Shade's approach turns file systems into data systems. Media becomes queryable. Search becomes semantic. Manual tagging becomes automated.

For teams drowning in digital assets, this matters. The file you need exists somewhere. The question is whether you can find it in five seconds or five hours.

AI-powered metadata generation closes that gap. Whether you're managing media libraries, documents, or any other unstructured content, the pattern is the same: automate the tagging, make everything searchable, then build workflows on top.

Shade makes the case that your file system should work more like a database. And once it does, you can finally treat content as the strategic asset it's supposed to be.