Unlocking Cloud Streaming's Full Potential: GPAC's ROI and Customization Power
What are the challenges when building your own live-streaming cloud?
Cloud services are convenient, but once you scale, you may realize that you're paying too much and you are not as flexible as you'd like to be.
In this article, I hope to convince you that the cost of customization that you have when using GPAC is actually an investment with a very interesting ROI, if you make the right choices.
First of all, what is GPAC? GPAC provides tools to process, inspect, package, stream playback, and interact with your media content. Many companies are faced with a difficult choice when deciding how to encode and package video. The main options are to go for a commercial solution, or whether to go open source, and to which extent?
However, many companies lacks the confidence of skills to fully implement GPAC, but it's possible, and it's easier than you might think. How do you decide what is the right solution for you, and what are the reasons why it's better to implement GPAC in your video streaming operations for transcoding and packaging?
We noticed that our customers face three major challenges.
The first challenge is related to features.
Of course, you need to know exactly what you need, and there are three main considerations.
I think the first to decide is about the capabilities, the codecs, the format, containers, DRMs, captions, and also how you plan to deal with metadata. Metadata is increasingly important.
The second thing is what we call “playing nice with the ecosystem”. You will need to interact with external players and analytics probes, ingesting specific contents with some specific protocols.
The third is about the features and extensibility. For example, you need to experiment with, let's say, low latency, HDR, or we need to consider interoperability. Sometimes you need to add custom flags to support some exotic devices, for example. Some solutions come with some nice graphical interface, but yet to handle all those codecs, all those parameters, all those specific use cases, you need to dig into some configuration details to make it work at scale. Netflix used GPAC, and they operate at the largest scale. I think there's no question here the GPAC supports many features. When it comes to features and innovation, GPAC takes roots in research and standardization. It is often one step ahead of the industry, but yet we cannot implement everything. That's why an open source software like GPAC also needs sponsorship and contributions from the industry. Another reason why GPAC plays nice with the ecosystem is that we are widely available. Mostly because of our efforts in standardization, people use GPAC as a reference implementation. Also, thanks to Motion Spell, GPAC is part of many plugin systems already available in the industry.
The second challenge is the important question of cost.
What is the cost, and how do you optimize it? Usually, we divide it into CapEx, which is the capital investment, the initial cost, and then OpEx, which is the operational running cost. You also have some other metrics, for example, such as environmental efficiency. GPAC is written in C, and C is known for its speed. As an open source software, it's being scrutinized by thousands of eyes. There's a good chance that GPAC is already as efficient as possible. If that's not the case, our community would complain about it, then make proposals to address any such issues.
If we agree that it is efficient and we can then focus on the money. Indeed, it's difficult to evaluate the cost, because each use case is different. There are a lot of features in GPAC, and each business comes with its own specific needs. You may ask yourself questions, such as: Do I need a CDN for all my channels? Do I need premium encoders for all my contents?
On the CapEx side, there is the cost of integration, and when using some open source software, integration is important. Some costs also are probably nontrivial to evaluate. For example, dealing with errors. When you're in operations, you have errors. The industry use case for this is the Synamedia Quortex architecture, which shows that errors can be handled smartly and efficiently.
A couple of months ago, I was invited by OTT Verse to compare the cost of the different options. There was a seminar called Buy versus Build versus Open Source, which provided a comparison of the different cost structures in different countries.
We took an example which was a broadcaster with an existing catalog of 100,000 content items and 500 concurrent users pulling them. The cost of the packager and doing all the packaging, the initial CapEx is $100,000 to $200,000, depending on the labor cost in your country.
The running cost is then almost negligible, compared at least to the transcoding cost. The cost of running open source means moderate CapEx with low OpEx. And maybe the interesting conclusion is that in all the use cases, open source ranks number one or was a close challenger in cost efficiency in all cases. For me, that's a confirmation that open source, when bundled like we do with GPAC and Motion Spell professional services, with professional and efficient encoding appliances, like the ones provided by NetInt, is perfectly aligned with today's challenge on efficiency.
The last challenge is flexibility, to make sure that you move fast in a fast-paced environment.
Let’s look at how Netflix successively went from SVOD to AVOD, from on demand to live, from AVC/H.264 to newer codecs, and from many containers to one container. All these moves, executed over short periods of time, contribute to making Netflix profitable. Again, I think that makes a difference.
Do you need to do the same? You can do the same. We are talking about open source software.
For people who are not aware, GPAC now provides a new application called gpac, that enables users to build completely custom media pipelines. Unlike historical applications of GPAC, which typically provided a fixed media pipeline, this new GPAC application allows for the creation of tailored pipelines to suit specific requirements. This includes transcoding, packaging, content protection, networking, and in general, any feature you need.
Let's move on to the first industry use case about the first challenge that was about being full-featured.
The first use case is Synamedia Quortex - a just-in-time “for everything” paradigm.
With their approach, the media pipeline is divided into small components that may fail, and may be relaunched when they fail. That's completely different from the traditional 24/7 transcoder that runs in a room, that should never fail, and requires a backup for seamless switching. Interestingly, the Quortex approach works incredibly well for live, at least for live up to a certain latency. For the latency that we usually deal with in the industry, it works really well. It's highly effective because it can run on any instance. You can take these churns and execute them on cloud spot instances that are cheaper (but that may be interrupted), or any available computer on your network. You can generate the content on demand. When a customer wants to watch specific content from an exotic device, you don't need to generate it prior to consumption. It's either in the cache already, or it is going to be generated just-in-time. That includes the packaging, but also the transcoding, and all kinds of media processing that would be needed. Eventually, Quortex was bought by Synamedia. The technology was made possible thanks to their vision and talented team, but also a strategic and continuous partnership with Motion Spell.
The second example is about the second challenge: cost.
Instagram, which is part of Facebook Meta Group, mentioned that they leveraged GPAC's MP4 box to alleviate video compute times by 94%. This choice prevented a shortage within 12 months of their capacity to provide video uploads for everyone. That's really a big change, and that's a really positive perspective that such big wins are available. This is particularly interesting. It just made more sense for them to just transmit and package most content instead of transcoding each content, because most Instagram content is watched only a few times. The cost of transcoding, in their case, outweighs the savings on the distribution cost. That's perfectly in line with the efficiency trend of the media tech industry. If you combine it with the Quortex just-in-time approach, where you could deliver content that is not very popular to a consumer encoder (such as your phone). You could use a professional encoder, for example a NETINT encoder, for content that requires better compression. For people who want to know the GPAC packaging computational complexity, it's close to a bare copy. That's a 94% savings in compute times, because the 6% remaining are mostly inputs and outputs, for example, transfers. Again, Instagram leverages this optimization because they control their system. That's why I think it's important, that's a big point of open source. Packaging in itself looks free (when done well) compared to transcoding. It is like copying/organizing the data by chunks without major modifications.You control everything, and it can make a big difference.
The last example is about the last challenge, flexibility.
People commented on the deal that we got with Netflix, and they asked if this was about money. This was not about money. We announced in 2022 that Motion Spell played a pivotal role in Netflix's successful integration of the GPAC software in its global content operations.
This integration has transformed Netflix's video encoding and packaging platform, resulting in an unparalleled streaming experience for millions worldwide. Netflix and Motion Spell, myself, went on stage at the NAB Show 2023 Streaming Summit. Here are some key elements. First, Netflix explained how the ISO BMFF, the MPEG ISO-based media file format, MP4, CMAF, is ubiquitous in their workflows. Of course, they are dependent on external providers, so ingested content may be in two main formats, as they explain. One of them is the progress formats, which is a bold move, which belongs to the ISO BMFF family. We have some good support available for this in GPAC.
Mainly, the recommendation of Netflix, for all the formats that they ingest, including the ads, is to go for MP4 or MOV. The Netflix streaming service uses CMAF. They also explained that their organization is totally aligned and committed to open standards and innovation.
By using GPAC, they got some innovations for free, because other people sponsored it, or because we implemented it before they actually needed it. Also, a company like Netflix sponsored some features of GPAC, such as the support of the xHE-AAC codec VP9, AV1 support, some improvements on the program's file format, etc.
Motion Spell collaboration with industry giants like Netflix is a testament to our commitment to innovation. As I mentioned, Netflix successfully went from SVOD to AVOD, from on-demand to live, from H264 to newer codecs. They executed fast and well. They initiated their container convergence, and they're going to show a lot more. That was made possible because GPAC is easy to integrate, GPAC is highly efficient in operations, and because it offers an incredible feature set, probably already beyond what you currently need, thanks to its root in research and standardization.
Most of GPAC is available in one executable, called GPAC. You can try it, deploy it, and get your service running. GPAC is cost-competitive because open source is free and easy to integrate. If you want to make the most of it, Motion Spell can help, as we did with Netflix and others. In general, we are always happy to hear about our users. GPAC only gets better thanks to your feedback. Lastly, GPAC's unparalleled flexibility will help you optimize and differentiate at a fast pace. Think Netflix interactive Bandersnatch, intelligent previews, outstanding captioning, and accessibility.
And finally, how can Motion Spell put this to work for you?
Motion Spell provides everything you need.
Motion Spell is the bridge between GPAC and its real world applications, and the driving force that enables organizations to fully harness GPAC's potential. The company's expertise extends beyond consulting, support, and training. As the exclusive commercial licensor of GPAC, Motion Spell provides a full range of professional services, ensuring that you can tap into GPAC's capabilities for your specific needs.
Feel free to reach out to Romain Bouqueau (firstname.lastname@example.org).
This article is based on the NetInt presentation, Deploying GPAC for Transcoding and Packaging, recorded on 17th August 2023.