AV2: What the Final Specification Means for Implementers

The Alliance for Open Media has reached a significant milestone in open video coding with the official release of the AV2 specification. While the announcement underscores major performance leaps and new capabilities, it also signals a critical shift from standard development to active implementation across the ecosystem.

To explore what this release means in practice and what lies ahead, we spoke with Andrey Norkin (Principal Architect, Video, NVIDIA) and Adrian Grange (Senior Technical Program Manager, Google), Co-Chairs of AOMedia’s Video Codec Working Group.

What does it mean that the AV2 specification is now “final”?

This initial release of the AV2 specification signifies that the coding tools are complete. It provides a stable foundation that implementers can use to immediately begin developing and finalizing their products.

While future versions are expected—including additional profiles and metadata to better support both existing and new use cases—the core coding tools will not change, ensuring long-term stability for developers.

What changed between earlier AV2 previews and this release?

The preview release was deliberately issued as a tool-complete version to give hardware implementers a head start on their design process before the final specification was locked in, thereby accelerating time to market. With those foundational designs underway, subsequent development focused on resolving open issues and refining the high-level syntax. This final release delivers the stable, comprehensive specification required to bring production-ready products across the finish line.

What does this enable for developers and implementers right now?

With the specification finalized, development can now advance across the entire ecosystem. This enables full-scale work on both decoders and encoders, encompassing both software and hardware implementations.

Software implementations typically represent the initial step, while hardware support—which is essential for widespread device adoption—follows. Reference software is already available, and efforts to develop optimized implementations are well underway.

AV2 delivers improved compression efficiency. What does that mean in practice?

Based on objective quality metrics, including VMAF and PSNR, AV2 delivers the same quality with approximately 30% better compression efficiency compared to AV1 across a wide range of content types and resolutions. Visual quality gains are expected to be even higher, reaching closer to 40% in bitrate savings. In practice, this means video providers can deliver the same visual quality using fewer bits, or significantly enhance quality while maintaining the same bitrate.

These improvements have direct benefits for both providers and viewers by reducing bandwidth requirements at scale and enhancing video quality for users, particularly in environments with limited connectivity.

The specification supports multi-stream and multi-view scenarios. How should we think about those capabilities?

AV2 enables multiple video streams to be combined into a single, synchronized bitstream, expanding how video content can be structured and delivered. This natively supports multi-view or composite video experiences where several related streams must be delivered together.

Furthermore, it facilitates advanced scenarios where the final video is composed from multiple components, such as texture plus depth, or background and foreground combined with an alpha channel.

Does AV2 introduce entirely new use cases, or improve existing ones?

Rather than introducing entirely new applications, AV2 primarily focuses on significantly improving existing ones. Enhancements in compression efficiency and stream flexibility directly translate to better performance across common use cases like streaming and video conferencing, positively impacting quality, responsiveness, and overall delivery efficiency.

That said, AV2 also natively supports more complex scenarios, such as multi-view video, which previously demanded complicated synchronization at the application level.

How will AV1 and AV2 coexist as adoption grows?

AV1 and AV2 are expected to coexist for a significant period. AV1 is already widely deployed across devices, and its adoption continues to expand. Meanwhile, AV2 adoption will grow gradually as implementations mature and hardware support becomes more broadly available, following the typical transition pattern observed in previous generations of video codecs.

What happens next following the release of the AV2 specification?

With the specification finalized, the focus shifts squarely to ecosystem implementation and adoption. This path involves the continued refinement of software implementations, followed by the development of broader hardware support. As these solutions mature, AV2 will increasingly be integrated into commercial products and services.