Meta adopts AV1 codec for real-time communication at scale

Meta's engineering team published a detailed account of its multi-year effort to bring the AV1 codec to real-time communication (RTC) at scale. The post spans the full journey: why Meta selected AV1, how it determines device eligibility (AV1 is computationally heavier than older codecs), rate control under variable network conditions, and error resilience for lossy connections. It shares the technologies Meta built to improve AV1 call quality and how it expanded coverage across its RTC stack powering calls in WhatsApp, Messenger, and Instagram.

While not an AI headline, the piece is relevant infrastructure: better, more efficient real-time video underpins the multimodal and live-AI experiences (video understanding, live agents) that every major vendor is racing toward. AV1's better compression at equal quality reduces bandwidth, which matters for streaming both human and AI-generated video.

The deep-dive is the kind of practitioner-grade engineering content that signals where Meta is investing beyond model headlines, and it contrasts with the company's rougher week on the AI-policy front (the paused keystroke-tracking training program and a 'AI for Work' product head's departure). For developers building RTC or video-AI products, the rate-control and error-resilience techniques are the actionable takeaways. Watch whether Meta open-sources any of the AV1 tooling.