What is video transcoding? (Insights for 2021)

video transcoding in 2021

Over the last year or so, a large chunk of the global population has been forced to stay at home and social interactions are happening over the internet. Most of it is via video content - with streaming video filling more than 50% of combined internet usage according to a recent report by Sandvine and up to 29.6% increase between February and March 2020 as seen by Fastly CDN.

Take a look into what the state-of-the-art in video transcoding is trending towards:

  • What is video transcoding?
  • How does video transcoding fit into developing a user-friendly content pipeline?
  • Current leading video formats.
  • New video encoding formats to look forward to: H.266 and AV1.

What is video transcoding?#

Let's trace the lifecycle of a video. As raw video data gets recorded by camera sensors, it is usually encoded into a video compression format since raw data can be quite large. For example, at relatively common sizes of 720p these days, a 24-bit 720p uncompressed video recorded at 60 fps clocks in at 1.32 Gbit/sec [1]. Encoding raw video is analogous to compressing raw RGB pixels of an image into formats like PNG, JPG, WEBP etc.

This is where transcoding comes in - simply put video transcoding is changing compressed digital video data between different formats (for example H.264 to VP9). Various competing video encoding standards evolved in lock step with different hardware device ecosystems. Changing bandwidth requirements as well as royalty/patent restrictions further shaped the adoption of these codecs. Certain compression formats also have hardware-enabled encoding/decoding support these days [2].

How does video transcoding fit into developing a user-friendly content pipeline?#

Generally, transcoding is a compute-intensive operation and setting up a user-generated video content pipeline required a large scale cluster of transcoding servers. The storage of various transcoded versions also requires a large amount of disk space. That's why most businesses that include any kind of video content on their platform rely on Infrastructure-as-a-service (IaaS) providers to reduce costs at scale as well as to simplify operational complexity.

If you want to add any kind of user-generated videos on your platform, you need to solve for 3 things:

1. Letting users upload video from their devices
2. Transcoding the uploaded videos in multiple web-ready format as well as sizes
3. Distributing transcoded videos to end-users via a CDN + video player implementation

Current leading video encoding formats.#

The leading video encoding formats broadly belong to two families: {H.264/AVC, H.265/HEVC} & {VP8, VP9} which are generally distributed in container formats like MP4 and WEBM respectively. Almost all the devices, browsers etc currently support one of these popular formats. On Youtube, you can right-click on a video and select "Stats for nerds" to see what encoding format they are using for the current video.

For example, here youtube is delivering a vp9 encoded video:

youtube stats vp9

Both VP9 and H.265 offer significant improvements over their predecessors but with the exponential increase in video-on-demand data consumption the media delivery industry is looking forward to the widespread adoption of the next generation video encoding standards.

New video encoding formats to look forward to: H.266 and AV1.#

As video content has become more popular everywhere, we've seen increased bandwidth load on internet infrastructure. These new formats aim at delivering high-quality video at lower bitrates.

If you're interested in knowing how video quality is measured, check out this article by netflix introducing VMAF. Some of the measurement tools rely purely on statistical models while others include percieved quality metrics as well.

AV1#

AV1 is the successor to VP9, developed by the Alliance for Open Media (AOM). Since AV1 has a royalty-free licensing model, it can be used in situations where H.264/265 family of encoders might not be usable.

In terms of performance, testing done by Facebook showed AV1 achieving more than 30% better compression than VP9, and gains of 50.3%, 46.2% and 34.0%, compared to x264 main profile, x264 high profile and libvpx-vp9. [3]

AV1 MSU performance stats

Source: Graphics & Media Lab Video Group, Moscow State University


Since AV1 is a newer codec, it is geared more towards high resolution data streams and is set to shine for 4K resolutions and beyond. One of the tricks used by AV1 to reduce bitrate requirements is called CDEF "in-loop constrained directional enhancement filter". It is an algorithm designed to reduce artifacts in the compressed video frames while also being easily vectorizable (ie is optimized for modern CPUs to run instructions in parallel on large batches of data).

H.266#

The H.266 codec is also known as VVC (Versatile Video Codec). It was developed by the Fraunhofer Heinrich Hertz Institute and finalized in July 2020. VVC supports a wide range of resolutions from SD to HD to Ultra-HD (8K) as well as 360° videos.


Similar to AV1, the work culminating into this new video encoder focuses on improving data compression but unlike AV1, VVC requires a paid license. Comparing bit-rates required to encode videos to the same percieved quality, VVC boasts a reduction of around 50%. According to Fraunhofer, a 90-minute 4K video encoded with H.265 (HEVC) clocks in at 10GB vs 5GB for VVC at the same quality. These newer codecs are especially well suited for mobile networks - fitting more video streams within existing data budgets.

The science#

Simplify your video pipelineTry MediaMachine today!

  • Get access to one of the cheapest Cloud-Transcode pipelines

  • Engage users early with great Thumbnails and NLP-Like Video summaries

  • No credit card required

Get started for free →