A vintage street scene in Los Angeles from 1947 shows a busy urban thoroughfare filled with classic cars and red trolley cars. The wide street is lined with multi-story buildings with prominent signs such as "J.C. Penney Co." and "HOLLYWOOD", under a pale sky.
AI media tagging that finds everything, instantly
Metadat-AI tags your video, images and audio automatically, then adds the metadata to your asset management and creative tools. Your team finds the right moment in seconds, not hours.
Your archive costs money until you can find it
Archives keep growing, but your team's time and budget do not. Editors lose hours hunting for clips, then re-shoot or re-buy content you already own. Every asset you cannot find costs you to store and earns nothing.
Manual tagging cannot keep up. Generic AI tags are noisy, inconsistent and expensive to run at scale.
What gets tagged
Every model runs on your media and writes its results back as searchable metadata. Open any model for examples and detail.
- Scenes and objects Setting, location, important objects, mood and lighting, described in plain language.Video and image models
- Actions What is happening, including sports actions such as a goal or a tackle.Video and image models
- Speech What is said, in which language and by whom, as a searchable transcript.Audio and speech models
- Text on screen Titles, captions, signs, sports scores and kit numbers.Video and image models
- Faces and people Recognise faces, with the right legal controls in place for GDPR and the EU AI Act, plus non-biometric sports player recognition.Face recognition
- Logos Recognise the brands that matter to you, such as sponsors.Video and image models
- Sports moments Goals, corners and cards, and non-biometric player ID for territories that ban it.Sports recognition
- Offensive content Over 50 categories and age-rating classification, to protect your brand and audience.Auto-Compliance
- AI-generated video A score for how likely each video is to be AI generated or manipulated, so it can be labelled.AI video detection
Transcripts are understood too: summaries, synopses and question answering (documents coming soon).
Search in plain English
Ask in plain English and search video, audio and images together (documents coming soon). One question, one search, every media type.
Works inside the tools you already use
Get the productivity gains of AI search in the apps and workflows your team already knows. Metadat-AI integrates with asset management systems such as Iconik, and we are an Adobe video solutions partner. Today, metadata reaches Adobe tools through your asset management system. There is no new tool to learn and no switching between systems.
Metadat-AI results inside Iconik: a scene description added as time-based metadata. Test clip from the open source film Meridian.
Pre-trained and custom models
Start with a ready-to-use model library. When you need something specific, such as your own products, people or logos, train a custom model with drag and drop and no data science.
Less noise, faster search
- 87% less metadata noise with Unique Frame ID, so results are more relevant. How it works
- 4.5 hours a week saved searching for assets, on average, by teams that save time with asset management (MediaValet 2025 DAM Trends Report).
- Simple per-hour pricing with no cloud egress fees.
- Up to half the price of comparable AI video platforms, with commitment and prepayment discounts.
Example output
Real output from the open source film Meridian. Unique Frame ID picks the frames that matter (the green marks on the timeline), and each one is described.
Scene descriptions
A street-level view shows the Earl Carroll Theatre, with large signage reading "EARL CARROLL THEATRE" and a slogan about "the most beautiful girls in the world". People walk along the sidewalk near passing cars on a sunny day, with a tall palm tree beside the building.
Classic cars are parked or driving in front of a large multi-story building labelled "Meridian". Pedestrians walk on the sidewalks, with streetlights, power lines and large arched windows framing the city.
Transcript
00:26.3JakeReid Scott
So these guys just disappeared?
00:30.6MacKevin Kilner
Yeah, on that stretch of road right above El Matador. You know it?
People
Mac (Kevin Kilner) and Jake (Reid Scott), recognised from reference images and voice recognition.
Your data, your environment
Run Metadat-AI in your own cloud or on your own hardware. With Incognito mode, we never receive your media or your metadata.
FAQ
What media types can it tag?
Video, images and audio today. Documents are coming soon.
Will it work with our asset management system?
It integrates with Iconik today. Other systems can be connected through our integration team or partners. We have well documented APIs, an API playground and LLM skills to make integrations easy.
Does it replace our current tags?
It adds metadata to your existing records. You stay in control of what is kept.
Do I have to use every model?
No. Choose the models you need. Pricing is simple and per hour.
Can the models run in my own environment?
Yes. That is how Metadat-AI is designed to run.
How long does a first demo take?
30 minutes or less. Bring a sample of your media and we will show you what Metadat-AI finds.