Create Unique, High-Performance Ads with #VideoAdAI

Video is a critical tool for any marketer that needs to break through the noise of the market and engage their audience with compelling content. In fact, marketers have consistently listed the use of online video to increase engagement as one of their top five opportunities for the last several years.

But the ways in which brands need to create and deliver video are rapidly changing—especially for advertisers. The explosion of social media channels and mobile devices mean there are more ways to watch video than ever before, and they each come with their own set of rules.

With so many options, it can be challenging to identify the networks, channels and keywords necessary to target the right audience. And when it comes to delivery, a single ad might need to be delivered in captioned or non-captioned versions, as well as multiple aspect ratios and resolutions depending on the delivery platform—everything from 4×3 to 16×9 or even 1×1 square formats. Attention spans are changing too. The traditional 30-second ad and two-minute corporate video now need to work alongside six-second, snackable ads or long-form, serial content designed to engage audiences over months-long campaigns.

Vishy Swaminathan, a principal scientist with Adobe Research wanted to see if he could use the power of AI to simplify the process of targeting and delivering video ads. He combined the ongoing work of three different Adobe Research projects into a single, integrated effort called #VideoAdAI and showcased the work as part of sneaks at Adobe Summit.

The first component of #VideoAdAI uses advanced research in AI and deep learning to understand and recognize the videos, and classify them to generate ad-related metatags. “We’ve seen other Sneaks capable of generating metadata for video, but this is a little bit different because it needed to be actionable for advertising rather than content management. It’s more about identifying demographic segments and subjects in the video that are relevant to the ad platform, so you can match the right ad with the right slot during bidding,” Vishy says.

The second component of #VideoAdAI uses a deep-learning neural network to analyze an ad’s content and compares it to the historical performance of thousands of other ads, then predicts how well the ad will perform with different audiences and delivery channels. “By looking at key performance indicators (KPIs) from previous campaigns and ads, we were able to teach the neural network what kind of content works best on specific channels like Instagram, Facebook or even TV,” Vishy explains.

#VideoAdAI automatically generates a 15 second video out of a longer version, for improved watchability on Instagram.

The third component of #VideoAdAI makes stylistic editing recommendations for improving an ad’s performance and automatically generates recommended edits, thumbnails and flashcards for promoting the video. “We showed editing for length in our demonstration, but we are working on other recommendations around stylistic preferences like . The generated edits can be imported directly into Premiere Pro and edited as necessary, but the great thing is that an advertiser doesn’t need to go back to the creative team for every single change or version if they’re already happy with the results,” says Vishy.

Vishy says the ultimate goal of #VideoAdAI is to help marketers and editors avoid repetitive editing tasks, get the maximum watchability out of their video content and close the feedback loop with the creator or editor of the video so that they have the final creative say.

“What excites me the most is the potential to use AI in a way that helps both the editor and the marketer, providing them with the information they need to make good decisions and maximize the creative and performative potential of their work,” Vishy adds. “The compute capability and access to data we have now is unprecedented, and it’s opened up a whole new world of creative applications for AI. The algorithms we use for AI and machine learning have been around for a long time, but we’re finally able to apply them in new ways to understand how audiences consume and respond to video as well as make the creative process more efficient.”

This story is part of a series that will give you a closer look at the people and technology that were showcased as part of Adobe Sneaks. Read other Peek Behind the Sneaks stories here.