SVG Sit-Down: Eluvio CEO Michelle Munson on Personalized Video Delivery for FAST Channels
The company’s Content Fabric aims for better monetization in free ad-supported streaming
Story Highlights
Eluvio’s Content Fabric demo at IBC last month showed how it can help take FAST (free ad-supported streaming) channel offerings to the next level by opening a whole new level of personalized ad insertion to create better monetization opportunities. The open and decentralized, streaming, content-distribution, and storage network is built for the third-generation internet, delivering live streams with deterministic end-to-end 2-second latencies globally to standard streaming clients (DASH/HLS over HTTP). It also provides a complete full-featured media stack to publish, store, and deliver content at scale, including personalization, access control, content protection, and proof of engagement. Eluvio co-founder/CEO Michelle Munson sat down with SVG to discuss Content Fabric and its role in the FAST-channel landscape.
Eluvio’s Michelle Munson: “The [main goal of Content Fabric] is content insertion that is data-driven.”
There are a few things that drive what we’re trying to do. One, it has to be very efficient because the margins are low. Two, you want to have a lot of channels with a lot of different ad insertions that drive real value to the user. You need to be able to scale what gets inserted along with personalization without trading one versus the other. Third is the profitability problem for broadcasters because they don’t currently control the ads, which are mostly coming from third-party platforms.
We want to address all those problems first by making it very efficient to have a lot of affordable FAST channels with low margins. The client needs to have server-side ad insertion that is personalized, because, [with] most FAST channels, ad insertion is coming from Google and Google is the one who makes money on that.
What is the challenge with live sports in the FAST environment?
With sports, there is a fourth challenge: the issue of bringing premium content into a FAST channel where there is no security in the legacy distribution model. In a decentralized model, we can do native watermarking directly in place, along with encryption, authorization, and only 2 seconds of end-to-end latency, which is very important.
At the World Cup last year, there were lots of good examples of FAST channels, but they required so much customization with standard players, and they also didn’t have ad capabilities. We want to replace the cloud stack, the media cloud, and the CDN with the unique aspects of the Content Fabric, which include low-latency security. With the Content Fabric, you no longer have the origin server or the CDN in the middle of the workflow: you have a uniquely watermarked output for each one of these clients.
Why is that important?
Our structure allows for avoiding file-based rendering and duplication in the traditional distribution supply chain. When the client requests an output, they are just operating with the source material without needing to render the file or do pre-transcode. If you need to add new audio tracks or something else, that process is completely componentized [without the need to replace the original content file]. That’s through the whole chain, not just the editorial phase: only the raw material needs to go through the network. That means 20-times less resource usage because you don’t need to remake content and store it in different variations. And it’s all in real time.
What was your focus at IBC?
At IBC, we demonstrated 30 feeds coming in from the Fox Technology Center in Tempe, AZ, as we wanted to show we are a serious player for the sports community. Our efforts the last few years are starting to pay off in that we have quality and efficiency coming together.
The demo showed that: we had Fox Weather channel feeds coming in, and the ad content was on the Fabric already and tagged with interest groups [for instance, people who like food or travel]. Those ads get stitched into the Fox Weather channel stream for the relevant consumer and in a way that is repeatable and open.
In the demo, we had three [viewer wallet addresses]: one in Phoenix, one in Los Angeles, and one in Washington, DC. They are each different addresses, and an access pass authorizes the Fox Weather channel in the wallet. The channel stream goes into the Fabric from Tempe, and we have an API function called POV Stream, which inserts the ad-content object. There is no external ad insertion needed: it’s all server side and personalized for each consumer. And the cost to do that is negligible because there is no additional resource required. And, because the content has data and the insertion is dynamic, you can have any business logic you wish.
Machine learning is a big topic at the show. Does that fit in as well?
For VOD, you can have machine learning tag the content library down to the frame level, making it a self-service platform. In the Fabric is a series of machine-learning models, including speech-to-text. You can run that over a large archive and have automatic labeling of the content for data-driven functions.
With generative AI, we can get full generative captions associated with the content and automatically generate clips from the source material. We have five models for that: celebrity detection, logo detection, landmark detection, speech-to-text, and optical character recognition (OCR). Those are the value-added features that make the Content Fabric a smart CDN.
The [main goal] is content insertion that is data-driven, because one of the problems with the legacy workflows is that you have to put your content into the cloud, get a data tag, and then pay to get it out.
