Per-title encoding with machine learning is the topic of thie video from MUX.
Nick Chadwick explains that rather than using the same set of parameters to encode every video, the smart money is to find the best balance of bitrate and resolution for each video. By analysing a large number of combinations of bitrate and resolution, Nick shows you can build what he calls a ‘convex hull’ when graphing against quality. This allows you to find the optimal settings.
Doing this en mass is difficult, and Nick spends some time looking at the different ways of implementing it. In the end, Nick and data scientist Ben Dodson built a system which optimses bitrate for each title using neural nets trained on data sets. This resulted in 84% of videos looking better using this method rather than a static ladder.
Optimising encoding by per-title encoding is very common nowadays, though per-scene is slowly pushing it aside. But with so many companies offering per-title encoding, how do we determine which way to turn?
Jan Ozer experimented with them, so we didn’t have to. Jan starts by explaining the principles of per-title encoding and giving an overview of the market. He then explains some of the ways in which it works including the importance of changing resolution as much as changing
As well as discussing the results, with Bitmovin being the winner, Jan explains ‘Capped CRF’ – how it works, how it differs from CBR & VBR and why it’s good.
Finally, we are left with some questions to ask when searching for our own per-title technology to solve the problem we have such as “can it adjust rung resolutions?”, “Can you apply traditional data rate controls?” amongst others.
MUX is a very pro-active company pushing forward streaming technology. At NAB 2019 they have announced Audience Adaptive Encoding which is offers encodes tailored to both your content but also the typical bitrate of your viewing demographic. Underpinning this technology is machine learning and their Per-title encoding technology which was released last year.
This talk with Nick Chadwick looks at what per-title encoding is, how you can work out which resolutions and bitrates to encode at and how to deliver this as a useful product.
Nick takes some time to explain MUX’s ‘convex hulls’ which give a shape to the content’s performance at different bitrates and helps visualise the optimum encoding parameters the content. Moreover we see that using this technique, we see some surprising circumstances when it makes sense to start at high resolutions, even for low bitrates.
Looking then at how to actually work out on a title-by-title basis, Nick explains the pros and cons of the different approaches going on to explain how MUX used machine learning to generate the model they created to make this work.
Finishing off with an extensive Q&A, this talk is a great overview on how to pick great encoding parameters, manually or otherwise.
Thursday 27th September 2018, 19:00 BST / 11am PT / 2pm ET
Encoding and transcoding are at the heart of every video service and solution, and the codec and format landscape has never been more crowded. Publishers are wringing the most efficiency out of H.264 while making the move to HEVC/H.265 and AV1—and keeping an eye on other proprietary codecs. On top of all that are considerations like video optimization, bitrate ladders, and per-title encoding.
Join this expert panel as they discuss the latest in encoding and transcoding, including the following:
The state of the art in encoding efficiency in 2018
How per-title encoding and machine learning can increase quality and decrease delivery costs
How to build flexible and cost-effective encoding solution
The latest developments in video encoding platforms and infrastructure
The benefits of contribution to distribution encoding and transcoding
The next big advances in encoding and transcoding, including AV1