During my research of metadata, I have run across some interesting discussions about it. The basic definition of metadata is “stuff that describes the thing,” but not all metadata is the same. Metadata can be created by those who make the video, or generated by other users in the course of a video’s life. In this article I am going to discuss some broad and specific metadata, with the help of two posts I have run across that I believe tell the story of metadata expertly. One is from 5 years ago at Udi’s Spot, the other is from 3 years ago at Philip Hodgetts’ blog.
Types of Video Metadata:
First, let’s talk about the kind of metadata you can actually control, or is automatic: the stuff that you personally write or your video equipment derives to describe a video. Philip Hodgetts has described about six different types of metadata, but as it got past three, some of them started to run together (much of the later ones require extrapolation from clues or logical exercise). So, I’m going to keep it short and leave it at three.
Hodgetts goes over these metadata types:
- Source Metadata. This is the type of metadata that is instantly created by the camera or editing software. File types, GPS, camera settings, time, duration, etc., are all technical details of a video that your production and post-production tools stamp on a video from the outset. In other words, it’s technical information generated by your various devices.
- Added Metadata. This is where you actually generate keywords on your own to describe the video, likely by hand during production and then entered into a keyword (tags, titles, descriptions) field during editing and publishing. Hodgetts also adds “comments” to the Added category, as a number of comments can increase visibility in search. These are all metadata that a computer can’t generate (at least not yet) on its own. While speech and facial recognition is getting better, it’s not entirely reliable as of yet. Even with “perfect” speech and facial recognition, a computer will be unlikely to be able to distinguish emotion or context in the dialogue. This is why you need to be able to describe your video to a computer in the best way possible. You can also add metadata during shooting (like the clackerboard) and editing (descriptions of takes) that will help communicate what each take is as you work towards a finished product.
- Derived Metadata. This is the area of speech and facial recognition, or GPS coordinates that “add up” to a certain place, contribute to metadata. Speech recognition can give you a transcript of the video, and facial recognition can tell you if the same person from another tagged piece of footage is the same person in a raw piece of video. Shot recognition can also tell you if the shot is a medium or close-up, as Final Cut Pro X does.
In the video SEO world, the user-generated metadata is the preferred kind. The words you use to describe the finished video and the user engagement dictates how a video will be located through a search engine. The other types of metadata are what producers and editors need to find the correct footage during the video’s completion.
Here’s a broad overview of metadata, as provided by the Udi’s Spot post. This is the metadata you basically have no control over:
- Explicit metadata, where users provide search engine fodder by rating a video or posting a video to a social media network.
- Implicit metadata, where it’s obvious that users have watched and enjoyed the video, without sharing it or ranking it.
In the world of video marketing, you kind of want both. What the Udi post argues is that one is more trustworthy than another: implicit metadata is an actual measure of relative enjoyment, while explicit metadata can easily be manipulated (views that are bought, or a video that is shoved through a social network with people who can manipulate the system).
You have some sort of control over explicit metadata, but it requires a bunch of work and you have to hope the right person finds the video and shares it. Implicit metadata, which you really have no control over at all, can be measured quite a bit from YouTube Analytics: how long are people watching the video, are they skipping ahead, are they turning it off before it ends? These kinds of things matter to a video’s overall life, and in YouTube’s database, it matters a great deal for search rankings.
The point of all this is that from shooting to editing to publishing, relevant metadata is important to increase video’s visibility. When we speak of metadata offhand on this site, we usually refer to search engine optimization. But there are many different types of metadata that increase a video’s value all throughout its life, and this is especially true if the video passes through a number of hands during the pre-published phase.