A short explainer (YouTube video) by Tsahi Levent-Levi discussing why AV1 — despite strong compression, especially for screen sharing and very low bitrates — isn't universally adopted yet. Main barriers include high CPU cost for encoding, sparse hardware encoder availability (expected to become widespread only over many years), platform optimizations still ongoing, and protocol-level engineering (decoders, SDP/RTP packetization, SVC negotiation). The speaker notes AV1's early WebRTC adoption (Chrome ~2020) and iterative improvements like switching decoders (libaom→dav1d).
Chapters
Why isn't everyone switching to AV1 video codec if it is so good? Well, that depends on so good for what I would say. Where we see AV1 being used and being tested at least are two areas. One is screen sharing because we know that screen content tools have good savings for slides. They're specialized for that job, so deliver huge improvements there. The other is very low bit rate. Okay, why not all scenarios then? The problem is that AV1 is much more powerful, but it also requires much more power from the CPU and we don't have the hardware encoders yet. That is only coming to the latest generation of GPUs. So you can probably rely on it being available generally in like 10 years if we're lucky. And the other thing is, what bitrate are we talking about? It's always a question. Is a codec good for the operating point we have in terms of bitrate? Yeah. So if I'm. If I have 2 Mbps then there are several problems that I have here. The first one is, let's say that I have 2 Mbps. I want to use AV1. AV1 will give me a higher quality, but if what I'm doing is 720p, I might not really need that additional quality. Yes. And I might not even be able to do AV1 at that point because I won't have enough CPU to use it for that bitrate. Yes. So it's kind of a trade off that we need to do here. Much more expensive in terms of CPU than OpenH 264 for example, which is still a software encoder. So I guess it's a matter of time for us to wait until the CPUs will have enough power. Not just the high end CPUs that we start buying like an iPhone 16, but rather all CPUs and have hardware encoders in them and be the majority of devices on the market. And the fact that we also. We didn't talk about that, but everyone is new, which means in particular on the Mac platform. Yes. So it needs to go through optimizations and figuring out what's best and how to use it and best practices and stuff like that that we haven't gone through yet with this codec as an industry. I mean it landed in Chrome for Web3C I think in 2020 and I think that was pretty fast how soon it was adopted there. But then it went through several phases like switching the decoder from Libaom to David as decoder and stuff like that. Lots of effort to be done, just new sdp packetization of rtp, figuring out how to do that. How do you do SVC with that? How do you negotiate these things? For example, we just had a change by a coworker of mine who changed the packetization and it got more efficient. It's turning to launch, so I guess it's a matter of time until everyone will switch to AV1. But it's not going to happen anytime soon. Okay, thank you. Thank you.
