To understand where AI and Integration play a key role at Indee we delve into the inner mechanisms of our product, while also exploring the technical intricacies of how some of the features are implemented.
The Indee product comprises four principal functions from an end-user perspective. As an enterprise customer, when you choose Indee as your video distribution system there are four stages your content goes through before it can be viewed by your intended audience.
Content ingestion
This is the process of uploading your content to Indee for processing. We achieve this through a collection of APIs that are referred to as the ingestion APIs. These APIs help create metadata about the content by creating the appropriate Indee entities such as ‘project’ and ‘video’.
At Indee, content ingestion is the first integration point.
When a customer uploads media files, our primary objective is to expedite this process, ensuring that content is uploaded within the shortest timeframe feasible.
To achieve this, we offer a highly scalable endpoint specifically designed to accommodate the swift and seamless upload of content. Furthermore, our platform facilitates multiple uploads concurrently, enabling numerous users to upload content simultaneously.
Consequently, the implementation of a highly scalable endpoint is the fundamental component of this process.
To realize this objective, we capitalize on the robust capabilities of our cloud provider, harnessing a multitude of resources to bolster the efficiency and scalability of content ingestion.
This content ingestion initiates automated processing, where we swiftly verify media compliance with Indee’s specifications by checking audio and video streams. If the media is determined to be fit for use, we promptly convert it to an internal standard format for downstream processing. Subsequently, Amazon Web Services (AWS) MediaConvert processes the media to a standard internal representation and storage.
Traditionally, we had custom web-based applications for our customers. As we onboard more technology-savvy customers, there is a clear ask for using ingestion in an automated fashion. Thus, we have launched the “Ingestion API” where the entire process of moving content and content-specific metadata to Indee can be directly integrated into the client’s system through a set of REST APIs provided by us.
Content processing and management
Once media has been ingested, Indee allows the creation of screeners. A screener represents an instance of the uploaded media that is created to be shared with a specific viewer. They can have specific rules such as expiration dates, maximum number of views, different levels of security, and so on. A collection of such screeners can be organized into various logical groups and presented to viewers.
A set of automated steps comes into play when content processing occurs. First, the video is converted to multiple bitrates to support adaptive streaming. This is followed by our core security feature, watermarking. Depending on the type of watermarking chosen Indee’s video subsystem can launch a few to several thousand servers to process various copies of the video to burn-in the appropriate watermarking patterns on the videos.
The challenges in the watermarking step are multifold. Launching multiple video processing servers is expensive and time-consuming. Our system optimizes both for cost and the completion time. It uses a priority-based queuing system to provision servers on the go and trigger the processing of the video.
During processing, the system is designed to accommodate various types of failures and requeues the videos dynamically based on the time available to complete the process.
Once the transcoding is completed, the media is moved to our content distribution network (CDN).
To enable our customers to create custom watermarking videos directly from their current workflow, our processing workflows are available as REST APIs as well.
This allows customers to determine the expiration date of the screener, the type of security, and the exact watermark content etc., for every individual video that they want to share with their customers.
This facilitates an intricate orchestration of video delivery from within client environments.
Managing audiences
Viewers in Indee can be grouped into distinct audiences and audience-specific content can be shared across the entire audience with additional metadata.
Audience Management allows customers to create specific campaigns targeting customer segments. The process involves creating an audience list of names and email addresses which can be uploaded to Indee. Clients can customize their content for each audience segment right down to the watermark. The Indee system then performs an operation that is similar to a “mail merge” and securely dispatches thousands of emails to recipients, each personalized with their screeners.
Our system also integrates with various email delivery systems to successfully deliver the email and track failures of delivery, if any.
Viewing content
Screeners on Indee are accessible on viewing apps, or through an API on the client’s custom viewing applications. Our native Roku, iOS and Android TV apps ensure that your content reaches audiences on all devices.
We provide customized branding for the content, multiple swim lanes for categorizing content and control over screener expiration dates.
We have also enhanced viewing experience and ease of access by incorporating OTT like features such as adding favorites, resuming playback from where they left off, etc.
To view a video, a multitude of systems and subsystems must work together. The client-side application, whether it is a web or native TV application, is a custom application developed by Indee for viewing authorized content. This application first makes a call to the Indee systems to obtain a list of authorized content and displays the same.
Once the user selects a particular media to be played, the application requests for the media manifest files (MPD/M3u8) for media playback.
The manifest files provide the DRM key information and the media chunks to be downloaded for playback. The player initiates the media chunk downloads from the CDN while making a request to the DRM provider for decryption keys.
During video playback, the system continuously monitors and records user engagement, including watched duration, pauses, forwards, and location via IP address and watch times.
This data forms the basis for watch analytics.
The intelligence layer
Across the four stages, there is an overarching intelligence layer that records the user behavior. This information forms an audit trail of key actions taken by the user on Indee systems. The audit trail covers both the end user and Indee support personnel.
The additional intelligence we are building on top of the audit trail is to discern periodic patterns in user behavior. Machine learning (ML) algorithms can discern such patterns easily and alert the project owners of any unexpected changes in behavioral patterns.
This forms the basis of our anomaly detection system – an intelligence layer that monitors user actions for significant change in patterns, raising alerts as necessary.
We see multiple use cases for such an anomaly detection system:
• Viewers: Changes in viewing behavior, including device switches, rapid IP address changes, altered viewing times, and content consumption speed, will be flagged.
• Administrators: The number of screeners being issued in a week, roles and their actions, number of videos uploaded, and so on can be monitored as well.
With an ML algorithm, we envision a greater understanding of customer behavior over time as well as providing a strong backbone for anomaly detection.
An API-enabled business
As part of our roadmap for 2024, we are looking at APIs enabling every stage of our system. The goal is to provide a seamless way by which customers can replace a particular stage with a custom solution and leverage existing applications and solutions within the Indee platform for managing their workflow.
Indee has been a high-touch business where our success has come from our customer success team. At the same time, an API model predicates that the customers build their own solutions and run and operate their solutions. While that may seem antithetical to one another, we see that dichotomy as being crucial to providing exceptional customer service to both business and technology teams.
We foresee customers picking a part of a stage or possibly one stage and building their own solutions, such as say ingestion and still leveraging the rest of Indee and our managed service offerings to complete their screening needs.
Our aim for the API-enabled business deepens our integration with customer systems and enables customers to build what they want faster.
================================
By Sharan Reddy, CEO, Founder, and Vivek Venugopalan, CTO, Indee