Russell Wise is Senior Vice President of Sales and Marketing at Digital Nirvana.


Media companies are in a race to create volumes of compelling, high-quality content for many different markets and distribution channels. But doing so has its challenges, explains Russell Wise.

Russell Wise is Senior Vice President of Sales and Marketing at Digital Nirvana.

Digital Nirvana

For one thing, the Covid-19 pandemic has turned the media industry on its head. Most media companies are struggling valiantly to produce content remotely at a time when in-person contact is restricted. That means remote production and cloud-based tools have become more important than ever.

The same is true for repurposing content. Even in the best of times, media companies have been keen to monetise their existing content. After all, some of these companies have vast asset repositories, and what better way to get more bang for the production buck than by using an asset in multiple ways, such as localising it for a new audience in another country? In this time of limited production options, repurposing content can be a critical source of revenue.

Then there’s the direct-to-consumer model, in which media companies avoid distribution points and go over the top directly to the viewer. To do it, they must be able to go through their libraries to sort, find, create, and distribute a piece of content, usually in four or five different versions, depending on the customer, geography, etc.

Even at the highest level, corporations like NBC are reassessing their strategies in light of the pandemic, deciding what media products they want to bring to the market and what technologies they need to make it happen. Those three scenarios – remote production, content repurposing, and direct-to-consumer delivery – are driving media companies to seek help creating compelling and compliant content more quickly, while adhering to government and internal standards and practices at the time of distribution. Most media companies are taking a serious look at AI and machine learning to help with this task.

There are clearly some discrete applications that are prime candidates for AI.

Content enrichment is one such example. At any time, you can add intelligence in the form of metadata; and at any point in the chain, you can produce better, more targetted content more quickly – and a lot more of it.

In pre-production and post-production workflows, a classic use case is to enhance the metadata of existing assets. AI can “watch” or “listen” and tag content in an existing library faster and more accurately than humans ever could. Or media companies can rely on AI to index any incoming feed so that metadata already exists before the content even enters the main system. In either case, having more – and more accurate – metadata speeds up the search process and accelerates content creation.

In one real-world example, a news website must deliver video within one hour of hitting post-production. The company uses speech-to-text engines to tag the video on its many incoming feeds, which makes it possible to meet deadlines and provide captions. Likewise, two major sports leagues are using AI for real-time captioning during live sports broadcasts. In all cases, AI helps broadcasters get content to air faster with better-quality captions – and without using human labour to do it.

Repurposing and localisation is another area where AI is critical. Once there is enriched content in the repository, it’s easy for media companies to repurpose it. For instance, they can use AI to translate content and localise it for other geographies. That’s exactly what one Spanish-language broadcaster is doing. AI takes Spanish content and captions it in different languages for different markets.

AI is also very useful where quality assurance is concerned. So, on the far end of the chain, for instance, there’s content distribution. This is where broadcasters check to see if the content is compliant to local government standards, like the FCC, or their own internal standards, like a style guide. They can use AI to assess, for example, the quality and accuracy of captions, which is becoming an issue in countries with stringent captioning laws.

Monitoring is another element worth considering. AI comes in handy when weeding out objectionable content, such as unacceptable language or images that would violate strict rules if they were to go on air. AI engines such as image and speech recognition can monitor and automatically flag such issues for review instead of a human continually viewing it. Likewise, AI can automatically quantify logo insertions, product placements, ads or even the number of times a given person appeared within a broadcast – all information that can correspond to billing.

AI technologies have improved substantially in the past years in terms of accuracy, but the big barrier to adoption in the media industry has been a practical application. That is, how to insert it into the content workflow. Take speech-to-text, for example. Broadcasters can get a lot of value from a speech-to-text engine, but it’s difficult for them just to order one. It has to come with management tools. It has to have a good UI and basic user functionality to get the full benefit. Fortunately, there are companies that have built a solid workflow that lets broadcasters harness the power of AI.

For instance, some products allow users to upload media through a portal to a cloud-based system. Once the content is there, the system essentially does all the work – transcription, translation, captioning, and monitoring. Users can set up presets to publish captions in the format required by the distributor, which is a pretty big thing, especially with Netflix. There are some basic tools for, say, editing the transcription, sort of like a word processor. There’s also a set of management tools that let users do things like assign, handoff, and track the progress of jobs. And when it’s time for content delivery, a content monitor automatically checks for compliance, quality and more.

The big benefits of AI in the content creation workflow are increased speed and reduced effort. The whole idea is to get the rudimentary, repetitive work away from humans. This way, humans can be creative while repetitive tasks can be designated to AI.