Every piece of content is different, and each requires specific encoding settings to look good. Since Netflix popularized the concept of flexible encoding ladders, Per-Title Encoding optimization has been gaining recognition for its quality and bandwidth-saving advantages.
It sounds simple enough, but there is a good reason it took the streaming giant years to make per-title encoding a viable part of their video delivery workflow. Each and every encode requires individual analysis as the first step, the process is both very time consuming and expensive (more about the reasons why here). Until now, implementing per-title encoding optimization was only accessible to the largest streaming organizations.
Bitmovin Product Manager, Daniel Hölbling-Inzko, explains the details in conversation with the Head of Developer Experience, Kieran Farr. The presentation covers:
+ The theories behind per-title encoding optimization
+ Real-life applications
+ Go through feature setup in the Bitmovin API.
3. Who are we? Daniel Hoelbling-Inzko
@tigraine
Product Management
Kieran Farr
@kfarr
Developer Marketing
4. Webinar Agenda
● What problem are we solving?
○ Diverse content libraries are challenging to encode!
● What are 3 common methods to solve this?
● Why is per-title encoding a good solution?
● How does per-title encoding work?
● Let’s look at some real world examples
● How to get started?
● Q&A
7. Problem
● Every piece of content is different
● So why are we using the same bitrate ladder for every video we encode?
8. Common Problems with Diverse Content
● One bitrate ladder is often far from optimal for certain titles
● High default bitrate is often wasteful (storage, CDN / bandwidth and
encoding)
● Quality is often inconsistent for consumers
10. Considerations
when choosing a
video delivery
strategy
● Bandwidth constraints of your customers
● Playback devices used by customers
● Kind of content you want to stream
○ Low vs high complexity
● Quality of source material
● Highest quality (bitrate) you want to deliver
● Switching between representations
○ Number of representations to encode
○ Smooth vs. clearly visible
● Cost of delivery
13. Recommended
General Purpose
Bitrate Ladders
But with this approach there are problems
● Action movies or sport movies would require
more bitrate
● Bandwidth is wasted for cartoons or news
● No consideration of
○ different aspect ratios
○ different FPS
15. Types of Content
● What are your different content categories?
● Who decides what content falls into which
category?
● Usually a manual process - Error prone
● Still a trade-off and not optimal
16. Every Content is
Different!
But they are still not optimal
● There are not only three categories of content!
Source: Netflix
19. Per-Title Encoding
● Compute a special purpose bitrate ladder for
every title
● Use bitrate to encode content in a quality that
viewers can actually enjoy, but not more!
● Improves quality, and reduce bitrate for less
complex content
● Fully configurable by user
○ Upper limit for highest bitrate
○ Lower limit for lowest bitrate
○ Step sizes between renditions (SD, HD,
UHD)
○ etc.
32. Comparison - Animated Content
● Bitrate Reduction of 28%
● Storage Reduction of 88%
● Quality Improvement 30%
33. Per-Title Encoding - Summary
● Every piece of content is different
● Per Title optimizes the bitrate ladder for each Asset
● Enables better quality of experience for customers
● Consistent visual quality across different assets
● Saves Encoding cost
● Saves Storage & CDN cost
34. How do I get started?
1) Contact sales to schedule consultation session (sales@bitmovin.com)
2) Activate this feature on my account
3) Change my API call based on analysis (provided after consultation)
Good morning everyone. Thanks for joining us for this webinar on per-title encoding today Thursday January 18.
Bitmovin is a provider of developer products to solve complex video problems. We have 3 primary products: a video player, video encoding and video performance analytics. Our software can be used through our API cloud offering, or on premise or in your own cloud using our containerized instances. We’re in 4 continents and growing quickly with top tier customers like the New York Times, Sling and Red Bull Media and the proud winner of a number of technology innovation awards.
First let’s start with introductions. My name is Kieran Farr, I help lead developer marketing here at Bitmovin.
Today our subject matter expert is Daniel Hoelbling-Inzko, a product management lead with deep experience in encoding.
Many of our customers have large video libraries with very different types of content. These frame show example of what we would call high complexity content. There is a great deal of movement in every part of the frame.
High amount of motion, and that motion is in most or all the parts of the frame.
Now compare what we just saw with these screenshots. Bojack Horseman is one of my favorite shows but let’s be real - it’s not the most visually complex TV show out there. Nearly half of this freeze frame is a solid color of blue representing the sky. Similarly with the “talking head” newsroom example while there is more complexity than just a blue sky, the background remains mostly still and the motion of the anchor host is relatively limited. These scenes clearly don’t contain as much data as those in the preceding slide.
So that brings us to our problem statement: every piece of content is different. So why do most publishers and platforms use the same encoding bitrate ladder settings for all content?
Lowest bitrate shows an enhancement in quality because it can benefit from a higher resolution.
Highest bitrates for both profiles look almost identical, although the bitrate of the standard profile is higher.