“Is Real-time AV1 Ready for Prime Time?” Webinar Session Two
As user expectations for technology advance, real-time communications (RTC) must keep pace. RTC faces distinct challenges, including ultra-low latency, adaptive network performance, and efficient coding for screen content. AV1’s suitability for RTC use cases including video calling, telehealth, and gaming, was addressed in part one of this two-part webinar series.
This second session, moderated by AOMedia Executive Director Pierre-Anthony Lemieux, was devoted to a panel discussion on AV1 and RTC and featured these experts:
- Michael Horowitz, Software Engineer, Google
- Sasikanth Bendapudi, Group Engineering Manager, Real Time Communication, Microsoft
- Tahani Trigui, Video & AI Software Engineer, Intel
- Patrick Ferriter, SVP Product & Marketing, Agora
Together they explored a variety of key considerations for using AV1 for RTC, a few of which are touched on below. To enjoy the full experience, we encourage you to watch session two on YouTube.
What Does “Ready for Prime Time” Mean?
Our experts talked about the need to meet the broad demands across industries and devices, and enable developers to deliver applications seamlessly across various hardware, including high- and low-end mobile devices, PCs, native apps, and browsers. Key factors include AV1's superior compression efficiency enables video calling on poor/low-capacity data networks and robust feature support that maintains network resiliency and adapts to fluctuating conditions without sacrificing quality. Compatibility with existing standards is also essential, allowing interoperability across systems. While there is still progress to be made, especially in global hardware support, a few major platforms have already implemented AV1 at scale, showing it can perform reliably and effectively for real-time communication at large.
Supporting Legacy Devices with a Low-Complexity AV1 Encoder
There are a few options for low-complexity AV1 encoders, including Libaom, which has been optimized to work on older Android devices without negatively impacting battery life, power consumption, or causing thermal throttling. This approach aims to ensure that even entry-level devices, some over five years old, can handle AV1 efficiently. Additionally, scalable complexity in codecs allows adaptation across devices, including low-performance options. Efforts like SVT-AV1 are also advancing this work, aiming to make AV1 broadly accessible across diverse hardware.
The Importance of Open Source
Open-source implementation has been crucial for AV1's development, especially for real-time communications (RTC), though it comes with challenges. With multiple large players contributing to the same codebase, coordination is complex and can slow progress, but it ultimately adds long-term value by encouraging shared improvements. Some companies, like Agora, began with an open-source AV1 library and then customized it extensively to meet their needs. Moving forward, a focus on interoperability within the ecosystem could accelerate adoption and usability. Supporting key ecosystem partners and refining open-source contributions to address specific use cases can also enhance AV1's impact.
The Promise of Machine Learning (ML)
For core RTC quality, ML can optimize factors like bandwidth estimation, improving real-time network adaptability, while advanced ML models could be used for encoding-specific performance gains. In the long term, there’s potential for a “neural codec” model that bypasses traditional hardware requirements, potentially simplifying processing demands across devices.
AI applications like generative AI also add practical features for RTC — such as transcription, real-time translation, and meeting summaries — enhancing usability rather than just improving encoding. Furthermore, ML can support visual quality evaluations in a more scalable way than traditional tools like peak signal-to-noise ratio (PSNR), offering automated assessments closer to subjective human evaluations.
Get the Full Picture
The full conversation with our experts goes into lots of great detail, so we encourage you to watch the full video on YouTube.