Encoding process

Is encoding a lever for decarbonization?

An initiative called "Greening of Streaming" sees this as the most promising way of reducing streaming's GHG emissions. At the same time, a publisher claims to have found a revolutionary way to reduce GHG emissions by 90% through "green" encoding. We bring you our point of view... and our decoding.

The “Greening of Streaming” collective has launched a campaign called “LESS” for “Low Energy Sustainable Streaming” to study how far the energy expenditure involved in video compression can be reduced to maintain acceptable quality. The problem is thus seen as one of energy sobriety, not digital sobriety.

“What if the ‘default’ streaming encoding profile was Energy Optimised with ‘acceptable’ quality for general viewing rather than, as it is today, Quality Optimised (and over provisioned) with no energy consideration?”

  • Would it be technically viable?
  • Would it really bring energy saving / sustainability benefits?
  • Would it be commercially viable?

It’s an interesting question. Let’s get back to basics.

How do you encode?

In simple terms… Video encoding (or compression) uses algorithms with numerous settings. Energy consumption and rendering quality depend on the machine time spent on calculations and the complexity of these calculations. If simple calculations are performed quickly, low-quality output is produced with little energy. Conversely, high quality requires a preliminary video analysis phase, followed by an optimized compression phase, and therefore higher energy consumption.

We always produce several qualities (between 4 and 6 in general) so that broadcasting adapts to the available Internet bandwidth. This is what we call adaptive streaming. The highest quality will never be better than the source video, and lower qualities are easier to produce.

In general, the computers that encode are of the same type as those that stream the videos. The energy consumption involved in encoding a video is comparable to that of a few playbacks. The more a video is played, the greater the impact of broadcasting than encoding.

Encoding can be hardware or software. It involves algorithms called codecs. Those that produce the best compromise between quality and bitrate are the most computationally – and therefore energy – intensive, both for encoding by a platform and for decoding by the user’s terminal.

Only one video codec is universal, supported by all Internet browsers on fixed or mobile terminals. This is H264 (also known as MPEG-4 AVC or H264 AVC). To encode in 4K/UHD or more, other codecs are more efficient, but they are not universal, and 4K is of no interest for fixed or mobile screens with lower resolution. In other words, except for SVOD services (Netflix, Disney+, Prime video…), encoding in H264 is perfectly sufficient.

To reduce the bitrate of a video, it’s simpler to restrict the quality available to the player than to create complex encodings (we call this “bandwidth throttling”). This makes it possible to reduce the amount of data transferred when playing older videos without having to re-encode them. It’s also a solution for backward compatibility and reducing obsolescence.

It should also be noted that resolution is not the only criterion for video quality. Frame rate (framerate) also counts. In a qualitative study, we demonstrated that a 720p video at 60 fps was mostly perceived as “better quality” than its 1080p version at 30 fps, despite a lower overall bitrate.

The bitrate of this 720p video at 60 fps is 5% lower than that of the same HD video at 30 fps, but its perceived quality is higher:

How do you judge the quality of an encoding?

A full HD image has 1920 x 1080 = 2,073,600 pixels.

One second of full HD video has 30 times that number, or 62.2 million.

The ratio of bits per pixel per second (BPPS) (or “Bits / Pixel x Frame”) is an indicator of the volume of data (bits) used to encode one second of video. To put it (very) simply, the higher the BPPS, the higher the encoding quality.

With the H264 codec, it’s considered unnecessary for the ratio to exceed 0.1, as quality gains beyond this level are imperceptible. Conversely, you should avoid going below 0.05 so that the compression is not too noticeable.

In other words, the encoding rate for a full HD video at 30 fps should be between 6,220 Kbps (62.2 Mpix x 0.1 bits/pix/s) and 3,110 Kbps.

The Avatar video above has a BPPS of 0.10, corresponding to a resolution of 1280 x 720, a frame rate of 60 fps and a video bitrate of 5895 Kbps. At the same BPPS, its bitrate would be 6220 Kbps in 1920 x 1080 at 30 fps.

Streamlike allows you to adapt encoding quality to the nature of the source. In other words, a ratio close to 0.1 will be applied for demanding videos (films, sports…) and close to 0.05 for cartoons or webinar recordings. The default quality is intermediate.

Should encoding quality be downgraded to reduce CO2 emissions?

An online encoding service claims to be “green” because it promises very high compression rates, which it tries to pass off as a reduction in GHG emissions! Of course, this makes no sense whatsoever.

We’ve tested this service and found that it performs highly degraded and indiscriminate encoding, whatever the nature or quality of the source. So much so, in fact, that repeated encoding systematically reduces the BPPS ratio to unacceptable levels.

If such a video is uploaded to YouTube, it will immediately be “upgraded” to a reasonable bitrate and BPPS, yet the quality cannot be improved.

This provides an answer to the question posed by “Greening of Streaming”: lowering encoding quality is pointless, as it is at the broadcasting level that the levers for reducing GHG emissions lie. Responsible streaming is above all a matter of common sense.

  • If a highly compressed video is uploaded to a social platform, its bitrate will be increased to comply with quality standards. Excessive compression is useless.
  • If a video is degraded during encoding, the lost quality is no longer recoverable, and the user experience is disappointing.
  • Conversely, if the delivery platform applies sobriety measures, the full quality of the video remains accessible, but it is possible to decide when to make it available (for example, only in full screen).
  • Sobriety is immediately effective, with no rebound effect. On the other hand, the benefits of a technological evolution are deferred and, above all, will encourage streaming more data at lower cost, which will worsen the environmental impact.

For a good understanding of bit rates and bits/pixel, here’s an old but accessible article:

Encoding is a far more complex issue than what we’ve simplified here. For more technical explanations, we recommend this site:

Share this post

Subscribe to this blog

Enter your email address to subscribe to this blog and receive notification of each new post by email.