Meta's David Ronca Talks Benchmarking and Deploying AV1 in the Android Ecosystem
The arc of AV1 codec adoption for large-scale content owners has been long and complex, as Meta Senior Media Software Engineer David Ronca (who also spent 12 years developing encoding solutions at Netflix) knows as well as anyone. In this interview with Jan Ozer of Streaming Media and Streaming Learning Center, Ronca discusses the integration of AV1 into the Android ecosystem, focusing on the challenges and solutions for mobile devices.
Catching up with Ozer after a presentation at the Mile-High Video conference, Ronca says, “Today I shared some of the work we've been doing to bring AV1 into the Android ecosystem, actually into the mobile ecosystem. But obviously it's the more challenging problem for Android devices, and that has involved a lot of benchmarking to understand the capabilities of the devices where we can and can’t deliver AV1 because we always need to ensure that we don’t give users a bad experience.”
Ronca goes on to explain that in the course of benchmarking AV1 integration into Android, Meta has developed VCAT (Video Codec Acid Tests), a new tool for benchmarking hardware and software decoders on Android devices. VCAT was designed to facilitate future benchmarking and ensure optimal user experience without compromising battery life.
Ramping Up AV1 with Software Decoding
Ronca also shares insights on the current deployment of AV1 and how they expect adoption to ramp up over time. “Fifty percent of video watch in the Meta family of apps ecosystem is now AV1,” he says, “and we’ve set a goal for the year to be above 60%. We believe that we can get it up to around 90% by January 2028.”
Ozer notes that the industry was predicting AV1 decoding in mobile phones by 2020, but hardware integration has progressed significantly more slowly. “Here we are in 2025 and maybe 10% of the Android ecosystem has hardware AV1 decoders,” Ronca agrees.
The biggest obstacle, he contends, are the economic constraints of hardware decoders in low-cost devices. “The reality is, just from an economic perspective, when you're selling an $80 phone, there's just not a lot of margin in that to put a bunch more transistors in the SoC [system on a chip] so that you have a hardware decoder. That’s forced us to rethink this,” he continues. “That's where we've leaned into software decoding, and we've been making investments in software decoding from several ways we’ve invested and talked about the work we've done for decode-friendly encoder recipes. You can actually reduce the complexity of decoding by making some better, more intelligent decisions on the encoder side without paying a price for efficiency or quality.”
As Meta continues to focus on making inroads to the Android ecosystem, he explains, "and get as many Android devices as possible into the AV1 pool,” more and more benchmarking has been required to reach an “understanding of what these devices can and cannot do, which ones we could target for AV1 and which ones we had to make sure we did not target for AV1.”
Extending Battery Life
Another key factor in advancing AV1, Ronca says, has grown out of work that Meta has done in collaboration with Intel and SVT-AV1 to “reduce decode complexity by 25%.” The catalyst, Ronca says, is “some better decision-making and more careful thought about how we use the various tools.” That 25% decrease in decode complexity, he adds, "came at about a 3-5% hit in quality. But when we deployed this, we simply used the next slower speed. So we bought that quality back with a little bit more compute.”
“At the end of the day,” he says, whether the method is conserving bits or reducing compute, the end goal with mobile device video is always “to see that work resulting in longer battery life. And that's what gave rise to the benchmark. Now we're running decode tests on Android devices--long-running battery drain tests--and we’re monitoring and tracking the battery. It’s a very difficult test because we want at least 16 hours of continuous playback. These phones now can get 20-24 hours of drain. So we realized that in order to do this level of benchmarking, we needed a tool to do this and hence the investment in VCAT.”
VCAT’s Benefits for Publishers
Regarding what this benchmarking tool enables publishers to do, Ronca explains, “If you think about the video distribution system" for mobile devices, "you have the network operators who manage bandwidth and network congestion. You have the content access provider (CAP) like Meta, Netflix, or others who are businesses built around delivering video. You have handset and SFC vendors, and we're all working on systems. And nowhere in there is anybody doing any testing or benchmarking to understand how these are going to be able to support the work that we're doing. In this case, [that means] getting AV1 out, but there's a similar problem for VVC where the hardware decode deployment cycle is going to be very long. We're used to short cycles from past codecs, but now the economics of the Android ecosystem don't support rapid refresh.
“So we're going to have to look at software," he continues. "And in order to look at software, we need to be able to look at these devices from their ability to support software. And by the way, they're getting very powerful, the modern ARM CPUs running as Android, all the advances in technology, they're making very, very capable devices. But nevertheless, we need to be able to measure that and understand that. So the VCAT tool, for example, would allow network operators to benchmark phones and determine how well they could support their given codec from a software or hardware perspective, and what battery life is going to look like for the handset vendors. If they're looking at SoC reference designs, they could also benchmark there. And with SoC, also, as they're looking at their next generation of chips, they could run benchmarks."
Ronca continues, "We think that a tool like this is going to allow the ecosystem from SoC all the way to the network operators and the CAPs to have a better understanding of the devices. That is essential to moving us forward."
What's more, he asserts, VCAT will make it possible for smalller content providers without the resources of a company like Meta to leverage benchmarking data before taking the plunge with AV1 on mobile devices. "If this analysis and benchmarking is being done broadly, then all the smaller companies would be able to leverage that work as well because the devices themselves would be benchmarked and rated, for example, possibly rated on battery life or whatever. So you could know that you could safely deliver AV1 to a device without having to worry about battery or thermal or any other problems."
“So if I'm a publisher and I have AV1,” Ozer asks, “am I going to distribute a different stream to different classes of devices, or am I going to send AVC to some devices? What decisions can I make at the streaming point to leverage this?”
“The more codecs you support, the more expensive your system is,” Ronca says. “If you start with AVC, and you put HEVC on top of that, let's say it's 40% fewer bits, but you're not reducing your bits at the edge. [So you're not decreasing] your cash costs and distribution costs by 40%; you're increasing them by 60% because you still have AVC and now you have HEVC and AV1 on top of it. Ideally, we want to get to a place where everybody has a legacy experience and then maybe one codec. That's where we want to be. When we get at 90%, we're literally saying to our users, 'Unless you're in a legacy experience, you would have AV1.'"
Ronca admits he has his doubts about reaching that ideal in the near term, but insists that it’s important to move toward it, considering the alternative. “If you have n codecs where n gets bigger and bigger every generation, you have to encode more videos, you have to push more videos through your network, you have to pay all that transit, and you have to store more at the edge. We'd like to see is newer codecs replacing the older codecs. The fact that you have to think today, 'Should I do H.264 or HEVC or VP9 or AV1?' It just gets too complicated."