AV2 vs AV1: What the Next-Gen Video Codec Brings to the Table
AV1 has become the go-to open, royalty-free video codec — powering streaming on Netflix, YouTube, and across modern browsers and GPUs. But its successor, AV2, is on the horizon. With the specification expected to land by late 2025, here’s what changes and why it matters.
Compression: 30–40% Better
The headline number is compelling. AV2 delivers roughly 30–40% bitrate savings over AV1 at equivalent quality. Internal benchmarks show a 28.63% reduction in PSNR-YUV and 32.59% in VMAF metrics. When combined with AI preprocessing, some reports push that to over 43% bandwidth reduction.
In practical terms: the same video, at the same quality, using significantly less data. That’s a big deal for mobile networks and CDN costs.
Bigger Blocks, Smarter Partitioning
AV2 introduces 256×256 superblocks (up from AV1’s 128×128) with fully recursive partitioning. This gives the encoder much more flexibility to isolate complex textures from flat regions, improving both quality and compression in a single pass.
AI Meets Traditional Encoding
AV2 isn’t a fully neural codec, but it takes a hybrid approach — integrating data-driven, trained models into the traditional block-based pipeline. These models handle tasks like intra-prediction more intelligently than handcrafted heuristics, without sacrificing the predictability and hardware-friendliness of a standards-based codec.
Motion Estimation Gets an Upgrade
Advanced optical flow techniques and a new Temporal Interpolation Prediction (TIP) mode give AV2 significantly better inter-frame prediction. Fast-motion content — sports, action sequences, gaming — benefits the most, with fewer artifacts and better detail retention.
Beyond Streaming: 8K, VR, and Screen Content
AV2 expands its reach well beyond traditional video streaming:
- 8K and HDR — unified exponential quantizer for 8-, 10-, and 12-bit depth, plus Wide Color Gamut support
- VR/AR and 3D video — native support for immersive content delivery
- Screen content coding — dramatically better encoding of desktops, text, and UI elements (a weak spot for AV1)
The Catch: Complexity
AV2 is more computationally expensive than AV1. Real-time software encoding will be challenging without dedicated hardware acceleration, and widespread hardware decoder support isn’t expected until 2027 or later. Early software playback is already being demonstrated — VLC 4.0 played AV2 content on a MacBook Pro at CES 2026 — but mass adoption will take time.
AV1 Isn’t Going Anywhere
AV1 is mature and well-supported. Netflix reports AV1 streams use one-third less bandwidth than H.264/HEVC with 45% fewer buffering events. Hardware decode is nearly ubiquitous on modern devices. AV1 will remain the dominant open codec for years while AV2 ramps up.
The Bottom Line
| Feature | AV1 | AV2 |
|---|---|---|
| Compression | Baseline | ~30–40% better |
| Max superblock | 128×128 | 256×256 |
| AI tools | No | Hybrid data-driven |
| Screen content | Limited | Dedicated tools |
| 8K / HDR / VR | Partial | Native support |
| Hardware decode | Widespread | Expected ~2027 |
| Royalty-free | Yes | Yes |
AV2 is a meaningful generational leap — not a revolution, but exactly the kind of steady, significant improvement that keeps open codecs competitive with proprietary alternatives like VVC/H.266. For now, build on AV1. Keep an eye on AV2.