What Is SVT-AV1 and How Was It Developed?

SVT-AV1 (Scalable Video Technology for AV1) is an open-source, production-grade video encoder designed to bring high-performance AV1 compression to multi-core processors. Developed through a strategic collaboration between Intel and the Alliance for Open Media (AOMedia), the project transformed AV1 from a computationally demanding experimental format into a practical, scalable standard ready for commercial streaming, video-on-demand, and real-time broadcasting.

What Is SVT-AV1?

SVT-AV1 is a software encoder and decoder library based on the AV1 video codec standard. AV1 was established by AOMedia to provide a royalty-free, next-generation video format capable of surpassing the compression efficiency of predecessors like H.264, VP9, and HEVC (H.265).

While AV1 offered superior visual fidelity at reduced bitrates, early implementations were resource-intensive and slow. SVT-AV1 solves this bottleneck by applying Intel’s Scalable Video Technology architecture to AV1. This architecture optimizes parallel processing across modern multi-core Central Processing Units (CPUs), allowing workloads to distribute evenly across threads and processor sockets. As a result, SVT-AV1 delivers high compression efficiency without requiring prohibitive amounts of time or specialized hardware accelerators.

Origins: Intel’s Scalable Video Technology

Prior to working on AV1, Intel engineered the Scalable Video Technology (SVT) framework to improve software encoding for other formats, notably developing SVT-HEVC and SVT-VP9. The primary objective was to harness the deep parallelism of server-grade processors, such as Intel Xeon CPUs, by dividing encoding tasks across multiple dimensions: segments, pictures, tiles, and blocks.

When AOMedia published the AV1 bitstream specification in 2018, the consortium’s original reference software, libaom, served as an algorithmic proving ground rather than a performance-optimized tool. Encoding a single high-definition video using libaom was excessively slow for real-world video workflows. Recognizing the need for an enterprise-ready software pipeline, Intel adapted its SVT design principles to the AV1 specification and open-sourced the early iterations of SVT-AV1 in early 2019.

The Joint Development with AOMedia

AOMedia—whose founding members include Google, Netflix, Amazon, Meta, Microsoft, and Intel—quickly recognized the promise of Intel’s scalable implementation. In late 2020, AOMedia officially adopted SVT-AV1 as its primary open-source software encoder, merging ecosystem resources to accelerate its development.

The joint effort combined Intel’s low-level hardware optimizations and threading expertise with the algorithmic innovations of AOMedia's broader engineering base:

• Hardware Optimization: Intel engineers contributed deep x86 optimizations, incorporating Advanced Vector Extensions (AVX2 and AVX-512) to accelerate computationally heavy mathematical operations.
• Algorithmic Tuning: Video experts from companies like Google, Netflix, and Meta contributed psychovisual tuning, rate control improvements, and advanced search algorithms.
• Unified Codebase: Community developers from independent projects (such as the SVT-AV1-PSY fork) and the broader open-source community added features, bug fixes, and broader architectural support, including ARM/NEON optimizations.

In early 2022, the project achieved a major milestone with the release of SVT-AV1 version 1.0.0. This release signaled full production readiness, offering stable APIs, a broad range of speed presets (from archival quality to fast real-time encoding), and performance metrics that made AV1 deployment economically feasible for cloud providers and individual creators alike.

Today, SVT-AV1 serves as the standard software foundation for AV1 adoption across the digital media industry, demonstrating how semiconductor manufacturers and software consortia can collaborate to establish sustainable open standards.