TV advertising is in the midst of a total transformation, driven by the prevalence of online streaming services that have led to a shift in consumer behavior and a new era of content consumption. Importantly, these streaming services are delivered by internet-connected TVs (CTVs), which provide for the first time the ability for TV content and advertising to be measured in a closed-loop system that enables the measurement of business outcomes that are associated with ad exposure, just like digital advertising. And as if this wasn't enough change for the TV industry, enter artificial intelligence (AI), which will accelerate the pace and scope of the TV industry’s transformation.  

For many decades, the TV industry has been relatively stable, with around $70 billion in revenue driven mostly by 500 or so major national advertisers making large ad buys based on “reach and frequency” and designed to drive awareness against target consumer demographics. The recent shift towards streaming services like Hulu, Roku and Peacock, coupled with the rise of connected TVs, has made TV just as measurable as online search and social advertising. New self-serve TV platforms have democratized access to TV, enabling millions of search and social performance advertisers to move into TV as a growth-driving medium. This shift is underway and will have a radical impact on the TV ecosystem.

AI will change the game yet again. With the help of AI, advertisers will be able to create TV-ready video ads by simply providing text inputs. AI algorithms will enable marketers to refine the ad and create multiple versions of the same ad with element-level differences, from major changes like an actor’s gender, color scheme or messaging, to more subtle alterations like frame rate, aspect ratio, types of clothing worn, volume, etc. Advertisers will be able to test multiple versions of ads on TV-buying platforms that provide feedback on exposure-to-outcome, to determine which combinations of creative elements drive the best performance (where performance equals outcomes like sales or return on advertising spend, or ROAS).

Moreover, TV buying and optimization platforms will use AI to optimize ads for outcomes, leveraging all versions of ad creatives in combination with an endless set of non-creative variables (such as context, audience profiles, daypart, frequency, etc.) to drive optimal results.  

Ultimately, creative development platforms will merge with TV buying and optimization platforms, enabling marketers to create TV ads with one-click simplicity, and then leverage AI systems to optimize creative and media buying using an endless set of variables.  This optimization approach, executed at the nexus of creative and real-time/programmatic TV buying, will transform TV from a reach and awareness medium into a powerful outcome-based medium, leveraging the unique TV attributes — sight, sound, emotion via large-screen, uncluttered delivery — to drive ads with higher impact and better ROI than any other advertising medium.  

AI is already accelerating the transformation of TV.  Within the next few years, TV will be accessible by millions of advertisers via democratized TV buying consoles that will be able to create ads, then deploy and optimize them using AI. The consequence will be a new TV advertising landscape that proves its value in a wholly measurable way.

Ever see an actor and wonder who he or she is or where you've seen them before? Or did you want to find a specific family member or friend in your personal video collection?

Real Networks claims that its two facial recognition products, being launched today (June 3), can find the face, using the Secure Accurate Facial Recognition (SAFR) software developed by a subsidiary unit. Initially the visual search technology will be available for shows on Netflix and YouTube; Real Networks Founder/CEO Rob Glaser told Multichannel News that the company is exploring relationships with other streaming video service to enable the Real browser plug-ins to support "as many platforms as possible." He declined to identify the next target systems or a timetable.

The new cloud-based products are StarSearch, which can identify thousands of actors in films and TV shows and provide pop-up information about them, and RealPlayer 20/20 to identify videos of personal friends/family.

An additional new feature can identify public figures who are in the news. Real's Discover web destination indexes popular news videos across the web, providing up-to-date information on the current events. It allows users to explore content through people in the news, as exemplified by this link to an MSNBC program on Tuesday night (June 2).

StarSearch is a free web extension that instantly identifies celebrities appearing on Netflix and YouTube programs and can pinpoint the scenes where celebrities appear in movies and TV shows. RealPlayer 20/20 is a redesign of the company's media player which uses a similar facial recognition algorithm to identify individuals, thus enabling consumers to browse, search, organize, and watch downloaded or local videos from their media collection.

According to Real Networks' research, more than 80% of consumers frequently ask themselves, "Who’s on the screen they’re watching" and often turn to a web search to figure out who it is.

StarSearch identifies famous people with one-click, using a redesigned RealPlayer user interface and downloader. The information and links appear on the viewing screen. The updated AI system enables viewers to organize video libraries by people, "a wholly intuitive way to search and experience content in your media library," according to Real executives.

Real Networks' facial recognition system is the latest in a series of efforts to use AI to help viewers do customized personal visual searches. Amazon’s X-Ray, available on Prime Video, similarly lets viewers get bios, filmographies, facts, trivia, character backstories, photo galleries, bonus video content by recognizing images during a program. X-Ray only works with Amazon Prime content - not on other platforms.

Related: FTC Issues Guidelines for Facial Recognition

Real Networks says it developed StarSearch and RealPlayer 20/20 as it was seeking to productize its SAFR research, which has been underway for several years. The computer vision technology scans and indexes videos for faces, and Real applied it to create a celebrity database, which automatically tags and labels personalities found in videos thus making video libraries searchable.

The company acknowledges that the search feature may be more valuable on YouTube and other services which feature programs with long-established stars with extensive on-screen credentials. Since many of the programs on Netflix, Amazon Prime and other streaming services feature young performers, the library of performances may be less extensive, hence with fewer links to video database references.

For Real Player 2020, the system will recognize faces and list them altogether; users will have to identify and label the family member or friend on at least one image, and the AI will then match the face to other videos in the collection; the systems will gather all the videos in which that person appears.

The company says that both systems can process factors such as age changes, which would enable to find movies with a "younger" and "older" performances of the same actor or videos of a child as a kid or a grown-up son or daughter. Glaser said it could also identify, for example, a bearded or clean shaven young or older actor such as Brad Pitt.

Converging Video and Data Applications

Real Networks declined to identify how many celebrity images are in its initial database - and the split between familiar stars and interesting character actors. The search process starts with clicking on a video frame that includes a visible face, which triggers StarSearch and RealPlayer 20/20 to identify famous people with a name, photo, current age, and "fun facts."

Real Networks says that its SAFR AI technology "has one of the lowest rates of bias" for skin tone and gender, according to National Institute of Standards and Technology tests of more than 100 algorithms. It also notes that SAFR has "99.87% accuracy [and] is three to five times faster than competing facial recognition algorithms, according to a July 2019 NIST analysis. 

SAFR can detect a face moving through live video, and then recognizes it, in under 100 milliseconds.

StarSearch by Real is available as a free web extension for Mac and PC through the Google Chrome store. RealPlayer 20/20 for Windows is available for free in the US and Canada, with plans to expand to other regions later this year.