AV1 vs HEVC: Should You Abandon HEVC Now?
For content distributors, the efficiency and cost of codes is essential to video editing. H.264 has been around for fifteen years. H.265/HEVC is still in the process of replacing H.264/AVC, yet the war of codec between AV1 and HEVC is already waged on.
AV1 (AOMedia Video 1) is an open, royalty-free video encoding format of the next generation. Developed by AOM (Alliance of Open Media Video), it is designed to be a successor of Google's VP9 and a competitor of H.265/HEVC. AV1 aims to improve coding efficiency by 30% on the basis of VP9/HEVC.
Main Features of AV1
1. Interoperable and open.
2. Optimized for the Internet.
3. Suitable for hardware with low computing power.
4. Adaptable to any bandwidth, able to be extended to any modern devices.
5. Consistent, able to make real-time video transmission with the highest quality.
6. Completely free, flexible commercial and non-commercial content, including contents generated by users.
AV1 vs. HEVC/H.265
1. Patent Licensing
AV1 is completely free, whether you are using it for commercial and non-commercial purposes. All members of AOM support AV1, an open-source, free ecosystem.
While for H.265, now it has a relatively high patent threshold. Four patent pools are related to it, they are, HEVC Advance, MPEG LA, Velos Media, and Technicolor in respective.
2. Hardware Requirements
Applying more advanced algorithms than HEVC, AV1 codes have higher requirements on the hardware. In other words, a processing chip with superior encoding performance is needed.
3. Compatibility
Theoretically, AV1 adapts the network better. It is designed as a video coding standard mainly created for the Internet and to be used in more network terminals. Though Apple's HEVC was announced in July 2017 to support HTTP Live Streaming, Apple itself has joined AMO's AV1 in 2018. Besides, royalty free patents entails a wide spread and evolution of the codec, since it involves more companies in play, customers would worry less about compatibilities issues.
However, in its initial stage, AV1 software decoders released by major browsers at this time are much slower and inefficient. For the time being, HEVC is more widely used with hardware support from Nvidia, Intel, AMD, Qualcomm, etc.
4. Container Format
At the time of writing, AV1 rolling out on YouTube test lists are using ISO Base Media File Format. Besides, WebM will probably be another choice for AV1, though the containerization was yet specified. Once standardized, AV1 would be widely compatible, as WebM format were developed by Microsoft, Google, Mozilla and Opera in a spirit of open design. Besides, AV1 also has Facebook, Google, Netflix and Amazon on its boat.
5. Compression Efficiency -AV1 vs. HEVC vs. VP9
We have made comparisons among the reference implementations using these coding standards. The reference implementation using AV1 is Libaom, while Libvpx for VP9, and x265 for HEVC. The test environment is AWCY, a public test set that contains more than 30 test video files from 360p to 1080p fixed to QP 60 frames. In such test environment, a conclusion can be drawn. Libaom has about 30% increase in compress efficiency over Libvpx, and 27% increase over x265.
Does YouTube Support AV1?
Though the most recommend upload video format for YouTube is still mp4, this video sharing giant has already established an AV1 Launch Playlist in 2018. You can check the codecs for a YouTube video by right click on the player > Stats for nerds, and look at the Codecs sections. If you see av01 something, that's AV1.
If you saw VP9 or other format in the AV1 Playlist, go to TestTube and switch from Auto to Always Prefer AV1.
Just like HEVC is an efficient replacement for AVC, AV1 takes a step forward and promises even higher compression without significant quality loss. YouTube's tagline for AV1 on TestTube echoes the benefit: more video for less data.
Final Words
Though there are still works to be done before AV1 goes widely applied, but the next generation of video entertainment is on the way to change our digital life dramatically.
