AI has the power to increase video accessibility, improve search safety and efficiency and package content intelligently says Pete Mastin, IBM Watson Media Product Marketing and Strategy, IBM.
To sketch a brief history of online video, consider the years following YouTube’s launch in 2005 as the first revolution.
Before then, online video was a novelty squeezed through a choppy internet connection, but now—a decade later—it is everywhere, all the time.
If the first revolution involved expanding video’s digital footprint, the second one, which is upon us now, is all about deepening that footprint.
Tech innovations that are upending our lives, including artificial intelligence, promise to create a two-way street between viewer and creator that simply wasn’t possible in the past.
From the ability to search video clips by image to powerful (and automatic) captioning, here are four ways AI technology will revolutionise what it means to create, find, share and watch video:
Enhanced Search and Discovery
The proliferation of online video content means more choices for consumers. But there’s a flipside: more video content also means viewers have more options to wade through to find what they want.
Video is classic “dark data,” and AI-powered video enrichment tools can make that data searchable.
The tools analyse the textual, audial and visual content of videos to recognise context and identify elements like objects, written text, brands and human faces. That information is extracted as metadata without human interaction and can be made searchable, directing viewers to specific clips.
By making video searchable, media companies can improve content discovery, increase operational efficiency, deliver higher ad revenues and improve viewer engagement.
“What do you get when you add up all of the elements AI introduces to video? Ultimately, you get content created by AI.”
Smarter Filtering and Compliance
Discoverability isn’t the only challenge AI will ease. It will also allow people to control access to video content that’s now easier to find. That should be welcome news to parents and providers alike.
Anyone who has used parental control features on their TV or mobile device has found themselves accidentally locked out of certain content, or had children given unfettered access to mature content.
That’s partly because the totality of a video can be ambiguous, with child-appropriate parts during some segments and adult content in others.
Detecting adult content, violence, objectionable language (in multiple languages) and racial insensitivity are serious concerns for providers and consumers. AI-enabled enrichment enables content providers to flag entire videos, or specific video segments.
This allows providers to easily check whether their content is aligned with communications standards or meets compliance regulations.
For viewers, it means software could do the hard work of determining whether video content is suitable for kids.
Closed captioning is the law. AI can improve outdated closed captioning methods, which currently fall into one of two buckets: accurate but laboriously created by humans, or automated and dodgy.
Compliance is challenging for companies that must caption diverse video content, including live news and weather broadcasts, sporting events and even archival content.
Changing closed captioning standards and regional legal variations add to the challenge. Too often, captions are either out of sync with the on-screen dialogue or only partially available.
AI-enabled captioning uses a multi-modal methodology, drafting instant—and accurate—captions.
Speech-to-text, natural language understanding, visual cues and other trained models dramatically improve automatic captioning. The result? Captioning that accurately reflect what’s happening on screen, whether a viewer’s watching a sports highlight or scrolling through their social feeds.
Facebook users watch more than 85 percent of videos without sound. AI-powered captioning opens new possibilities by enriching the stream of videos on social media—and perhaps even encourage people to stop scrolling, put on their headphones, and turn on the sound.
Smarter Way to Package and Promote Content
What do you get when you add up all of the elements AI introduces to video? Ultimately, you get content created by AI.
Don’t expect to watch a full-length AI-made feature any time soon. But the same technologies that can identify a face can identify the pace and tone of a video, and edit it according to moviemakers’ tastes, taking the manual labor out of creating trailers.
AI can spin out a working, screen-ready preview with ease. And instead of requiring an entire team of video pros, all that’s needed is software.
In fact, it’s happened already: Last year, IBM’s Watson created a trailer for the studio’s AI thriller “Morgan.”
Watson, having watched the movie and trained on the structure of trailers, produced an appropriately eerie preview with the cuts, stops, and cadence familiar to any moviegoer.
The trailer raises philosophical questions, even from those involved in its creation. “Will computers be able to create art? Can they create period?” John Smith, a fellow in machine vision at IBM Research said of the trailer.
Regardless of where you fall on that philosophical question, it’s undeniable that AI is a powerful tool to augment human creativity and make the content we create more compelling.
Pete Mastin is IBM Watson Media Product Marketing and Strategy at IBM.