AV1 Video Codec: Development History, Technical Architecture, and Cloud Encoding Applications

AV1 is the first‑generation video coding standard developed by the Alliance for Open Media (AOM). Since its release it has attracted great industry attention, and Tencent Multimedia Lab has joined AOM to help optimize the encoder and provide high‑performance cloud encoding services.

The international video coding landscape includes four major standard bodies: MPEG (under ISO/IEC), VCEG (under ITU), China’s AVS, and AOMedia. MPEG and VCEG have historically collaborated on standards such as MPEG‑2 and H.264/AVC. AVS represents China’s own standards (AVS1‑3). AOMedia, founded in 2015, introduced the royalty‑free AV1 standard, with Tencent becoming a board member in 2019.

AV1’s technical framework is a hybrid coding architecture composed of many modules:

(1) Block Partitioning – Images are divided into 128×128 largest coding units (LCU) which can be recursively split into smaller blocks using up to nine partitioning modes.

(2) Intra Prediction – Removes spatial redundancy using directional prediction, recursive filtering, Paeth operator, cross‑component prediction, DC and smooth modes.

(3) Inter Prediction – Exploits temporal redundancy with affine motion models, overlapped block motion compensation, and mixed prediction modes.

(4) Transform – Supports DCT, ADST, IDT, Flip‑ADST and up to 16 row‑column transform combinations.

(5) Entropy Coding – Uses a multi‑symbol context‑adaptive binary arithmetic coding engine, improving throughput over traditional binary arithmetic coding.

(6) Loop Filtering – Includes de‑blocking, constrained directional enhancement, and loop restoration (Wiener and self‑guided projection filters).

(7) Palette Mode – Index‑based coding for screen‑content video with sparse color values.

(8) Intra Block Matching – Example: Tencent’s logo uses vector prediction to reduce redundancy.

Key application scenarios for AV1 are streaming (YouTube, Netflix, etc.), live and on‑demand cloud transcoding (Tencent Video Cloud, AWS Elemental MediaConvert), and emerging use cases such as 8K video, VR, and even medical imaging where low distortion is critical.

The Q&A section explains why multiple standards exist (historical fragmentation of MPEG and VCEG), the advantages of AV1 (≈20% bitrate reduction over previous generations, royalty‑free licensing), challenges with patent pools, current hardware support (MediaTek Dimensity 1000, Samsung and LG 8K TVs), and Tencent Multimedia Lab’s role in standardization, patent contributions, and open‑source encoder development (SVT‑AV1, AV2 research).

Future directions include the next‑generation Versatile Video Coding (VVC/H.266) standard, which promises ~35% bitrate savings over HEVC, and AOMedia’s upcoming AV2 standard, with collaborative research between Tencent and Google.