Dav2d
VideoLAN releases dav2d, a fast open-source decoder for AV2, the new royalty-free video codec succeeding AV1.
dav2d is an open-source AV2 decoder in development by VideoLAN, released under a BSD-style license. AV2, finalized by the Alliance for Open Media, offers roughly 25% compression gains over AV1 but requires approximately five times more decoding complexity. The dav2d project builds on lessons from dav1d, the widely-deployed AV1 decoder used in VLC, FFmpeg, Firefox, Chrome, Safari, and Windows. The current codebase includes a feature-complete AVM v15 decoder supporting 8-bit and 10-bit decoding, with most major components implemented: bitstream parsing, entropy decoding, intra and inter prediction, transforms, filtering, and film grain synthesis. Performance optimization work is underway for x86 (AVX2), ARM (AArch64 NEON and arm32), and RISC-V architectures. The project benefits from checkasm, a validation and benchmarking framework developed during dav1d's lifecycle, allowing faster and safer optimization than dav1d's early stages.
What HN community is saying
Patent concerns dominate discussion. Top comments note AV1 faced patent claims from Sisvel and Dolby despite royalty-free designation, though Dolby's lawsuit remains unresolved and Sisvel has not successfully enforced claims in court against non-payers. The technical trade-off sparks debate: AV2 achieves 25% compression gains at a five-fold decoding complexity cost, raising questions about real-world performance and adoption of software decoding on existing hardware. Several commenters defend dav1d's actual deployment success (Netflix, YouTube, older devices like iPhone 7) against claims of niche status, though one notes TV performance degradation on AV1 content. C and assembly remain appropriate for this project given SIMD optimization requirements and fine-grained instruction control Rust makes painful.