AI auto tagging comes to ShareBrowser
At NAB 2018 we previewed a new feature in our product line, and, not surprisingly, it captured the attention of many people.
With just the click of a button we can automatically tag your videos in ShareBrowser. Meaning, you choose a video, click an icon, and within a matter of minutes the content in that video is automatically identified using artificial intelligence, tagged with keywords, and searchable.
Now, admittedly, it seems like everywhere you turn someone is using terms like “machine learning” and “AI,” and those phrases are already becoming passé before their real potential hits home. But when you experience first-hand what can be done with AI and you apply just a little imagination… everything else seems old.
Media creators will love AI
I’ve been fortunate enough to see a lot of new technology long before it hits the streets. I’ve held tiny pieces of hardware worth more than a Ferrari, and I’ve tested some of the most exotic storage components ever produced.
As much as I love technology, I’ve definitely built up some immunity to what impresses me. And as you might have guessed by now, I think the implications of AI will be huge for our industry.
I know we’re all walking around with more computing power in our pockets than what was used to go to the moon. But, let this sink in: You can now have a computer scan a clip and quite accurately identify the contents of that video. Almost instantly. It’s now good enough, easy enough, and inexpensive enough that it’s applicable to just about any organization.
Here’s a very short demonstration of what we previewed to our visitors at NAB:
In the ShareBrowser demo above you can see that we processed a short video of a ballerina. What comes back are very accurate keywords, including “Ballerina,” “Ballet,” “Dance,” and “Performer,” all of which get automatically entered into the ShareBrowser database.
Imagine sending thousands of hours of archived video through AI and having it quickly, accurately, and inexpensively, automatically processed and categorized so that it’s instantly searchable by your post-production team.
This could be one of the greatest things ever for curating B-roll footage!
Imagine running any new video straight through a first-pass auto tagging job once it hits the storage system. And better: having keyword-based rules and automations that fire whenever the AI finds, say, any clip with a convertible sports car.
AI for sports, news, marketing, worship
At NAB, we talked with customers who brought up interesting AI use cases for professional sports, advertising, news, worship, and product recognition.
Consider a workflow where professional athletes are automatically identified, by name, and their presence noted at exact points in the timeline. Maybe there’s a need to find an interview of an athlete, but more importantly—only videos where that person is wearing a specific team’s jersey…
Maybe you’re in a newsroom and want to find all videos that contain fire, or floods, or cats, or people carrying guns. Or maybe you only want a video that contains all of those things!
Or perhaps you’ll have a need to identify faces of people who aren’t exactly “famous.” You’ll be able to train the AI to recognize that person, and you’ll be able to assign other metadata to that person, such as their age, gender, etc. Putting that all together means a marketing team could search their library and easily find all video of, say, female executives only. A church could assemble a collection of recorded services including video of a certain guest pastor.
And the more we use AI the better it’s going to get. The video you submit to AI now will have even better recognition a couple years later. But of course the great benefit to all of this is that it won’t require anywhere near the enormous time, effort, planning, and cost that would be necessary for humans to do these jobs, leaving us with more time to do the best work that only humans are good at doing.
Interestingly, you’ll first use AI to find things you want, and then you’ll also use it to exclude assets that are no longer important. From there, you’ll almost certainly want to push your libraries through AI every now and then, but you’ll only have to send the parts of your library that have been found (with the help of AI) to really matter.
You’ll soon be able to click a video, effortlessly have it identified, tagged, transcribed and translated, and acted upon, all with AI. (The AI can also identify text in video, and it’ll probably even be able to identify brand/product logos, and eventually read lips.) You’ll be able to search all of it. And it will only cost pennies per minute of video.
I remember seeing 2Gb Fibre Channel and thinking it was impressive. Those days are long behind us, and I couldn’t be happier about that.