Internet video content represents a huge slice of online traffic (approx. 64%) and its quality is always improving, yet contrary to logic you don’t always need a faster broadband speed to keep up. In fact today a slower connection can view content that yesterday would have required something much faster.
According to Cisco, some 64% (21,624 PetaBytes per year) of consumer Internet traffic is video content (e.g. YouTube, Netflix, NOW TV, iPlayer etc.) and this could reach as high as 80% by 2019. Similarly the recent introduction of Ultra High Definition 4K (UHD 3840 x 2160 pixels) content is only adding to the capacity pressures being faced by ISPs and that’s before we even consider 8K.
However such forecasts often overlook innovations in compression and encoding technology which, thanks to the ever increasing power of modern computer processors (CPU), are enabling us to fit much more detailed video content into increasingly small packages and that in turn requires less broadband speed to deliver onto your screen.
The easiest way to explain how dramatic this change has been is to demonstrate with two examples of a bog standard High Definition (HD) video, which I recorded at 720p (1280×720) for around 43 minutes. The video itself is of a holiday event and contains lots of motion and colour (a sample screengrab is below – converted to JPG for the website).
By default the video actually took up around 5 GigaBytes of storage space because it was saved in a raw lossless quality format, but nobody uses that for streaming video online as it’s unnecessary and would be too expensive (capacity) to run, not to mention that many people would struggle due to the need for a faster broadband connection.
Instead I encoded the video, which ran at around 25fps (frames per second), using both the standard H.264 format and the more modern H.265 (HVEC). Both are widely used today and a lot of online streaming providers make use of similar technologies (e.g. YouTube uses VP9, which is similar to H.265). For a fair comparison I also tweaked the settings to make sure that the picture and audio quality was virtually identical. The resulting file sizes were as follows.
H.264 /MPEG4 720p [MKV]
Video File Size: 987 MegaBytes
Time to download at 2Mbps = 1 Hour 9 Minutes
Time to download at 24Mbps = 5 Minutes 44 Seconds
H.265 / HEVC MPEG-H 720p [MKV]
Video File Size: 243 MegaBytes
Time to download at 2Mbps = 17 Minutes
Time to download at 24Mbps = 1 Minute 24 Seconds
NOTE: We could have compressed the H.265 copy even more and made it just 160MB in size, but this would have lost just a little too much quality.
The outcome means that you could stream the H.265 encoded video live over a stable broadband connection speed of around 1Mbps (Megabits per second). Meanwhile the H.264 video of roughly the same length and quality would require a speed of a bit over 3Mbps. Mind you every video is different and some sections will demand a higher speed than others, so your experiences will vary.
Naturally for those on poor broadband lines, where sub-2Mbps connections can still be a problem, the advantage of using a modern video compression method like H.265 should be clear. This is of course the reason why many modern video streaming services look so much better today than they did a few years ago and all without you needing to upgrade your Internet connection.
Some of the earliest forms of digital video were produced by simply stringing together each frame (picture) one after the other in an approach that is not unlike how analogue film projectors worked, albeit done digitally on computers.
The problem with this raw approach is that it consumes vast amounts of storage space (GigaBytes) and, as demonstrated above, big files take much longer to download or require a faster connection speed to view in real-time over the Internet (streaming). This can be reduced by adding more compression, but that hurts the image quality and so is not ideal.
Thankfully modern computers and Smartphones’ have some incredibly sophisticated processors (CPU) inside. As CPUs improved then video compression systems have similarly grown more intelligent, not least in their ability to analyse, predict and then remove / ignore unnecessary content; let’s call this “better” rather than “more” compression.
For example, let’s say you’re watching a video of a kitchen with two people in where the camera records from the same position for 5 minutes and the only things moving are the people. Modern compression methods are smart enough to know that they don’t need to store the static parts of all the frames that go into the sequence, so they largely ignore the repetitive elements that don’t change and only focus on those that do.
The result is that you still see a fully intact video, but the compression only needs to store those elements in the sequence that change and the rest are reconstituted / repeated from the earlier frame(s). By taking this approach the file size is significantly reduced because you only need to store part of each frame / picture in the sequence.
Obviously this becomes much more challenging for scenes with a lot of moving and complicated imagery, such as video of a fast running river or heavy rain, but there are ways to tackle those too through improvements like motion compensation and spatial prediction.
Funnily enough H.264 did a lot of these things already, while H.265 (HEVC) improved upon them and found more efficient ways to produce roughly the same result. For example, H.265 found it more efficient to examine what does / does not change between frames by analysing larger sections of any given scene (coding tree blocks).
Going forward we might well see further improvements with future codecs, although there are some very real constraints on compression technology and as such it’s difficult to judge how much more room there is for this sort of enhancement. You can only predict and compress so much before running into the problem of diminishing returns. Mind you we’re told that the universe did at one point exist as a single infinitesimally small volume, so who knows.
Otherwise it’s thanks to enhancements like those above that you can today stream 4K video content online with a connection speed of around 20-30Mbps (Megabits per second), yet only a few short years ago the same video would have required 50Mbps. Go back even further and a 100-200Mbps connection would have been required etc.
At this point we should say that not every online video streaming service is currently able to make full use of these upgrades, which is partly because some older hardware would struggle to cope with the newest video codecs (requires more processing power) or they simply don’t support them, but that will improve with time.