MAM moves on through AI

  • With Kevin Hilton

MAM moves on through AI

In the second part of InBroadcast's look at media asset management in broadcasting, Kevin Hilton hears how artificial intelligence is offering more functionality for tracking and logging material in today's content-heavy industry...

There was a time when media asset management (MAM) in broadcasting could have been seen as ancillary to the main activities of production, post-production, distribution and archiving. It was necessary, of course, but more a support function in many ways. Today, with full digital, file-based working over multiple networks, MAM is even more of a necessity, being integral to the smooth running of broadcast operations.

This importance is reflected in the growth of MAM as a technology market. In 2021 the global sector was valued at $1.37 billion. The Fortune Business Insights survey of the MAM market published this year projects it will grow from $1.53 billion to $3.81 billion by 2029.

Key trends include an increasing demand for cloud-based asset management, which is coupled with the ongoing segmentation between virtual and on-premises working. Most significant of all, however, is the influence of artificial intelligence (AI) on how assets are managed and logged.

The integration of AI with MAM systems is already happening and playing a major role in expanding this once niche technology market. But it will go further, according to the survey, with AI being integrated into software "to enable production and quality assurance of training data within a familiar MAM environment." As well as this, users will also be able to recognise information in video or images files to create their own AI models.

Among recent moves in this direction was the May announcement by Arvato Systems that its VidiNet cloud-based media services platform, part of the Vidispine portfolio, was now working with DeepVA's AI software. The foundation of Vidispine's MAM system is provided by VidiCore, which, together with VidiNet, now gives users access to DeepVA AI through the VidiNet Cognitive Services interface. Arvato says the aim of this is to simplify media workflows over the whole system and allow people to use AI "within their familiar MAM environment".

A similar approach has been taken by Tedial, which introduced its latest system, smartWork, at the NAB Show in April. The company's Chief Technology Officer, Julián Fernández-Campón, comments that Tedial was looking to take a "wider approach" with smartWork, which is a NoCode media integration platform that allows AI tools to be incorporated into its MAM facilities. "More important, it is able to decouple the integration from the MAM," he says. "Users can configure and trigger AI analysis from different vendors with a few clicks, without having to know the internal details of each integration."

Fernández-Campón observes that AI capabilities "have improved significantly in recent years", which will continue in the future: "This has provided remarkable results in many areas. In some instances, AI has been used as an art generator. Also, customers have been asking for many years for AI to be used in some specific use cases such as celebrity identification and detection of inappropriate content, such as sex and violence.

It will certainly continue to be adopted in MAMs, where the key is to be able to evaluate and test AIs for each business case and content type as the results might not always be as good as expected."

At VSN, Product Management Director Toni Vilalta says broadcasters will be able to use AI to boost the efficiency and reliability of processes that are currently carried out manually. "This technology opens the door to automatic cataloguing of content in MAM systems, among many other functions. We can provide an enormous number of accurate metadata within seconds and users will only need to validate them, avoiding a manual and lengthy process that is prone to human error."

Vilalta comments that AI is something "present" for VSN and not in the future. At the moment VSNExplorer is able to integrate with the leading AI engines on the market, including IBM Watson, Google Cloud, Microsoft Azure and AWS. "These provide automatic metadata detection and cataloguing," he says. "Currently we are developing projects on AI with broadcasters such as RTVE to provide an automated catalogue of archive content. Our MAM system provides access to all information processed and extracted in a single window called Databinder. RTVE is using this technique to automatically catalogue its archive using the Etiqmedia AI engine."

For video, AI systems can be used to implement facial and scene recognition, while on the audio side it is possible to enable a full text transcription to be retrieved in a matter of minutes. "This is with capitalisation and accentuation," says Vilalta. "It also recognises people, places, events, products, organisations and dates. Furthermore, it extracts an automated catalogue of the content and main keywords." 

A further application of speech-to-text technology is highlighted by Bill Admans, Chairman of The Advisory Board of Ateliere Creative Technologies, who explains that audio can be scanned to create subtitles and captions. "The AI also translates into other languages, enabling content owners to inexpensively create highly accurate, localised subtitles and captions without the need for expensive human interpreters," he says. "MAMs are all about managing metadata and describing the content. Traditionally, metadata was technical data but now it is used to describe the content. New AI technologies can scan files and recognise what is happening in the content to create a library of metadata that can be searched and used by content creators."

All of this, Admans observes, is part of the shift in scope of what MAM systems can do and are being used for. "Today's MAM platforms have become content hubs, interconnecting all parts of the creative process," he comments. "Our Connect platform automates and orchestrates workflows using AI and machine learning [ML] to drive efficiency and accuracy. As the industry continues to adopt cloud-native workflows, access to powerful computing and services is bringing greater integration of AI and ML. Adobe, Amazon AWS, Google, IBM and Microsoft Azure are continually developing AI and ML that third parties can build upon to create powerful and unique technologies to evolve how we work and create content."

Cloud storage and data backup technology providers such as Backblaze are witnessing the increasing exploitation of AL/ML for cloud-based broadcast workflows, where accurate cataloguing and tracking are now essential. "When working with hundreds of hours of footage, it takes a huge number of hours to just watch everything, let alone catalogue it properly," says Vice President of Sales Nilay Patel. "AI can eliminate the manual effort and find the gems from the outtakes and tomorrow's b-roll from today's shots. This is where SaaS [software as a service] MAM becomes essential. When a customer shoots video, they can immediately upload it to cloud storage. The SaaS MAM then finds the footage, catalogues it and sends it to AI for analysis. That analysis then lives in the MAM for future discovery."

As David Candler, Senior Director of Customer Solutions at AI tech company Veritone, comments, the media and entertainment (M&E) sector is recognising that AI can bring similar benefits to those being delivered by the cloud. "The world of M&E is now turning its attention to AI to help create operational efficiencies, engage with audiences and create new revenue streams," he says. "According to industry reports, the market size for AI in M&E globally is estimated to grow nearly $100 billion by 2030. A report by Gartner shows that more than 80 percent of the world's data is unstructured and growing by 30 to 60 percent per year. This is why MAM vendors and their customers need AI."

As with any new technology, there are different ways to implement AI, from bespoke models employing different vendor APIs, which can be expensive and time consuming to deploy, to more self-contained operating systems. Veritone has taken the latter approach with its aiWARE platform, which offers closed file and real-time input adapters, plus multiple AI engines. "AI is the answer," says Candler. "For broadcasters it offers many practical benefits, from simplifying content management workflows to powering real-time, high volume content analysis." Veritone aiWARE can be integrated with the company's DAM (digital asset management) system, Digital Media Hub, but is also able to work with a variety of third party DAM and MAM systems. This is the case with the installation at Italian Serie A football club Inter Milan. The club is using the Evolphin Zoom Media Asset Management platform in conjunction with aiWARE and AWS web services to manage its library of video, audio and images. There may be some in broadcasting and the world at large that may object to the word 'asset' being used to describe video and audio material but the proliferation of TV and streaming channels has only further increased the demand for programming, both new and archive. Tracking it during production and retrieving it from libraries is now imperative. As Mathieu Zarouk, Director of Product Marketing at Dalet, comments, there is a need to make finding content easy, no matter how old or unlogged it is. "End-users are looking for ease of use and a clear user-interface that allows them to get on with their tasks quickly and efficiently," he concludes. "This is where leveraging AI can bring time savings when it comes to adding metadata and logging. Even if it doesn't quite reach human accuracy yet, a lot of manual work can be eliminated."