AV2 v1.0 Starts the Codec Race Before Silicon Is Ready

AV2 v1.0 is the published final AOMedia Video 2 specification, a successor to AOMedia Video 1 that the Alliance for Open Media (AOMedia), the standards consortium behind the codec, lists as version 1.0.0 with matching AOMedia Video Model reference software and a May 28, 2026 release date. It targets lower bitrate streaming and better screen content support.

The practical story starts after the launch page. Developers now have a stable target. Streaming services, browser teams, chip vendors and player projects must decide how much bandwidth savings are worth another codec transition while AOMedia Video 1 (AV1, the earlier open video codec) is still moving through devices.

A Final Spec Before Consumer Support

The release gives implementers a stable bitstream target. AOMedia’s page says the specification outlines bitstream syntax, semantics and decoding processes required for conformance. It also says the January working draft labeled v13 is superseded by v1.0.0, a useful warning because early chatter around codecs often treats draft numbers like product versions.

That table is why the release matters even if no viewer sees a new menu option this week. The standards work has moved from research churn to implementation duty. Every group after this point is judged less by promises and more by test streams, decode errors, speed and battery use.

The Compression Claim Has Test Data

AOMedia’s release language says superior compression, but the better number sits in research. In a May 15, 2026 arXiv submission accepted for ICIP 2026, five video coding researchers compare AV2 v13.0 against an AV1 baseline using the AV2 Common Test Conditions in the AV2 compression performance paper.

The paper uses Bjøntegaard Delta Rate (BD-rate, a codec test metric that estimates bitrate change at similar quality). Lower is better. The caveat is that the paper tests v13.0, not the final v1.0.0 release, so the numbers are evidence for the engineering direction rather than a consumer product benchmark.

Those are not small codec numbers. For a streamer carrying millions of hours, even a smaller real-world gain can change storage bills, peering costs and video quality at congested network edges. For a user, the same gain shows up only when apps choose the format and devices can play it without burning the battery.

Software Gets the First Assignment

The AOMedia Video Model (AVM, the reference software used to match the specification) is the first tool developers can build around. The official AVM v1.0.0 source tree is public on GitHub, the code hosting service used by AOMedia for the release tag.

Reference code is usually written for correctness before speed. That is not a criticism. Without it, independent encoders and decoders have no shared answer key when a bitstream fails or a decoded frame looks wrong.

VideoLAN, the nonprofit behind VLC media player, is already working the other side of the problem. Its dav2d AV2 decoder project describes a fast, cross-platform decoder based on dav1d, with speed and correctness as its stated focus.

• Finish a complete C implementation that is portable before assembly tuning begins.
• Expose a usable application programming interface (API, the calls apps use to talk to the decoder).
• Port the decoder across desktop, mobile and less common processor families.
• Add architecture-specific assembly for faster playback once correctness is in place.
• Improve threading so multi-core processors can help with difficult streams.

The list sounds dull until a player has to open a file on a laptop with no dedicated decoder. That bridge between standards document and video player is where codec launches either gain patience or lose developers. AV1 adoption benefited from dav1d because a fast CPU decoder gave browsers and apps a fallback before every device had perfect hardware support.

Hardware Will Set the Pace

AV1 is the useful memory here. AOMedia’s March 2018 announcement for the AV1 public release included a bitstream specification, unoptimized experimental software encoder and decoder, reference streams and binding specifications. The release gave silicon teams a target; it did not instantly upgrade living room hardware.

That history makes the spec-to-device gap the main risk for the new codec. A fixed-function decoder in a chip can play complex streams with far less power than a general CPU, which matters most on phones, tablets, streaming sticks and TVs. Until those blocks exist in shipping products, software carries the demo load.

Silicon schedules also move slower than standards pages. Chip vendors need a stable bitstream before design freezes. Device makers need chips before product refreshes. Streaming companies need enough reachable devices before they encode catalogs at scale. The release starts that chain; it does not compress it.

AOMedia’s Member Math Raises the Stakes

AOMedia’s September 2025 survey gave the release a demand signal before the final file arrived. The group said 53% of surveyed members planned to adopt AV2 within 12 months of finalization, while 88% expected implementation within two years, according to AOMedia’s AV2 launch preview and member survey.

Treat that number with some skepticism. The respondents were members of the alliance building the technology, so their answers are not the same as a neutral survey of the whole device market. Still, the result tells chip, browser and streaming teams that their peers are at least budgeting attention.

The roster matters as much as the percentage. AOMedia identified its steering committee as Amazon, Apple, Cisco, Google, Intel, Meta, Microsoft, Mozilla, Netflix, NVIDIA, Samsung Electronics and Tencent. That list spans cloud delivery, browsers, operating systems, chips, phones, televisions, social video and subscription streaming.

Broad membership does not erase the coordination problem. Each company has a different trigger for adoption. Netflix may value bitrate savings in premium streams. Mozilla may care about open playback on the web. A chip vendor may wait until the feature can be sold across a full media block, not bolted on for one app.

The Streaming Payoff Depends on Patience

The use cases on AOMedia’s release page are revealing: streaming, broadcasting, real-time video conferencing, screen content, split-screen delivery and augmented reality or virtual reality applications. None of those markets pays for elegance in a spec. They pay when lower bitrates hold up under ugly motion, text, bad networks and battery limits.

For streaming services, the first decision is likely selective use rather than broad replacement. New codecs often begin with high-value clips, test buckets, premium resolutions, or platforms where the service controls both encode and playback paths. The open question is how quickly AV2 can move from lab encodes to those controlled deployments.

If dav2d and AVM turn the spec into reliable tooling before chip vendors lock their next media blocks, AV2 can enter services as a bandwidth decision rather than a science project. If that software layer stays thin, the v1.0 badge will sit where codec milestones sit first: in documentation, test clips and engineers’ issue trackers.