AWS Media Intelligence Solutions Bring AI/ML to a Wider Breadth of Media Workflows

Subscribe To Download This Insight

2Q 2021 | IN-6123

On March 23, 2021, AWS announced the availability of its AWS Media Intelligence Solutions (AWS MI) that brings together a combination of services to serve a range of Artificial Intelligence (AI) and Machine-Learning (ML) needs for media workflows across three initial target markets (media and entertainment [M&E], education, and ad tech) and that currently targets four primary use cases.

Registered users can unlock up to five pieces of premium content each month.

Log in or register to unlock this Insight.


AWS Launches AWS Media Intelligence Solutions


On March 23, 2021, AWS announced the availability of its AWS Media Intelligence Solutions (AWS MI) that brings together a combination of services to serve a range of Artificial Intelligence (AI) and Machine-Learning (ML) needs for media workflows across three initial target markets (media and entertainment [M&E], education, and ad tech) and that currently targets four primary use cases.

  • Search and Discovery: Supporting metadata creation, object/asset identification (e.g., logos, characters, people/faces, scenes, activities, etc.), and content categorization, including frame accurate and precise search functions
  • Subtitling and Localization: Speech to text (Amazon Transcribe) and translation services (Amazon Translate, which supports 120 language pairs) for localizing outputs
  • Compliance and Brand Safety: Image, video, and speech analysis to ensure content compliance rules and regulations, brand safety, and standards
  • Content Monetization: Categorizing and contextualizing content for targeted advertising/sponsorships and content/service recommendations.

In addition to Amazon Transcribe and Amazon Translate, AWS MI is underpinned by Amazon Rekognition and Amazon Comprehend. Amazon Rekognition handles the image and video analysis used in the metadata tagging and categorization of content while Amazon Comprehend is a natural-language-processing service to analyze text and unstructured data. AWS MI itself is offered through two avenues: through its partner network (18 partners listed in the announcement—14 technology and 4 consulting partners) or via AWS solutions using frameworks designed by AWS (Meda2Cloud, AWS Content Analysis, Elemental MediaTailor), although companies with deeper levels of AI/ML expertise and in-house data scientists can leverage Amazon SageMaker to prepare, build, train, and deploy their own ML models.

Within the current competitive landscape AWS MI represents a relatively wide breadth of coverage, with many companies offering more targeted solutions—companies such as Valossa, for example, specialize in areas targeted by Rekognition and compliance. There are also other companies and solutions that target different areas within the media workflow. Companies like Viaccess Orca, for example, use AI/ML to better characterize audience data down to the individual user level, and companies such as Synamedia and Nagra leverage AI to enhance the customer base’s value (including reducing churn) and to monitor video quality.

AI/ML Is Becoming Table Stakes


Three-plus years ago, AI/ML was a buzzword within the M&E market, but solutions that employed AI were often viewed more as a trial feature rather than as an essential component. Progressing to today’s market, AI/ML is now viewed as a critical and often a highlighted differentiable element of a video company’s solutions or platforms. AI is being used across a range of functions and throughout video workflow, from content generation to distribution and monetization. The four categories targeted by AWS MI encompass many of the key target applications within the media landscape, although there are other areas such as content generation (i.e., automatic creation of clips/trailers), video stream monitoring (video quality/performance), and additional applications within the ad tech market (e.g., translating audience data to individual users, monitoring ad performance, etc.) that have also quickly become key use cases for AI/ML.

The market however is not constrained to applications and services within the consumer market; corporations and brands are also levering media solutions backed by AI/ML to process, manage, and distribute content intended for marketing and internal distribution. These media solutions are particularly valuable to companies that do not have a native expertise in media and that would benefit significantly from these tools and automated processes. Video recordings of meetings and collaborative sessions will become increasingly important and common as more companies adopt and embrace a hybrid workforce between in-office and remote, and companies will need to make this information both searchable (i.e., assigning appropriate metadata) and accessible (i.e., having cloud storage and media asset management system[s] that reach across a company and its groups).

AI/ML Will Rule Digital Marketing and Ad Tech


While AI/ML is increasingly finding applications within M&E workflows, its role there may pale in comparison with the impact it will have in the digital advertising and ad tech markets. Ad tech companies are already using AI quite extensively to target ads (both on behavioral and contextual bases), identify users, track attribution and conversions, and optimize ad campaigns. The growing push for privacy, however, will catapult AI to the forefront. Apple’s changes to its Identifier for Advertisers (IDFA), for example, is expected to greatly diminish mobile marketers’ ability to track individual users as most are expected to opt out of IDFA tracking. Perhaps more significant is the impending depreciation of third-party cookies expected to hit Chrome in 2022. Chrome is the last (and largest) major browser to end support for third-party cookies. Google’s privacy sandbox and proposed Federated Learning of Cohorts (FLoC) could become Google’s new standard that could bring a host of considerations for the ad tech market, websites, marketers, and end users. Although associating a user as part of a large cohort of individuals with similar browsing habits (with safety precautions in place to avoid associating users/cohorts with sensitive browsing habits) may sound like an optimal way to deindividuate a user from their browsing history while allowing marketers to continue behavioral marketing, there are some potential pitfalls with FLoC.

Some of these pitfalls include fingerprinting, matches of first-party data/log-in to cohorts, and the impacts of opt-in/opt-out decisions by users and websites. These pitfalls and the potential remedies will invariably engender dramatic changes to how companies collect data and how ad tech and advertisers employ addressable advertising. Regardless of how this facet of the market moves forward (FLoC has recently entered trials in select countries), AI will play a critical role as the days of easy access to user data is approaching its sunset. Companies (like AWS) that have core competencies across AI/ML, e-commerce, advertising, and cloud will have a tremendous opportunity to tackle these hurdles for customers as they arise. AWS with its partner program can also make a significant impact on the market without being viewed as one of the advertising monoliths such as Google or Facebook. AWS MI is solidifying a solid beachhead into a growing M&E market, but we expect to see much more from AWS and AI/ML in the ad tech market in the coming years.  



Companies Mentioned