Anyone working in video knows that along with fast startup and avoiding stalls (buffering), video quality is a key factor in satisfying your audience. We at Bitmovin have recently introduced three new ways video streaming services can optimize their streams to boost video quality without using more data.
This presentation demonstrates how encoding innovations, like per-title encoding, multi-codec streaming, and per-scene adaptation, help reverse this trend and reduce bandwidth requirement for distributing the same, or better, quality video.
2. Who are we? Paul MacDougall
Solution Architect
Kieran Farr
@kfarr
Developer Marketing
3.
4. Webinar Agenda
● Registration Survey Results
● What problem are we solving?
● Obstacles
● Existing Solutions
● Next Generation Solutions
● How to get started?
● Q&A
5. Registration survey results
● Improving visual quality of videos = 10.5%
● Reducing bandwidth required to deliver the videos = 8.5%
● Both, improving quality and reducing bandwidth = 81%
6. What problem are we solving?
For broadcasters targeting mainstream adoption of online video, most users
have the expectation of the same QoE as linear TV. There are many obstacles:
● Bandwidth Limitations
● Device Fragmentation
● Limited Insight Data
● Diverse Content
● Net Neutrality Issues
7. Existing Solutions
● Big bitrate ladder
● Ask your customers to get a better connection or upgrade devices
● Negotiate special peering agreements
● Better implementation of existing codecs
10. ● Use the most efficient codec possible
● Match codecs to browsers dynamically
● Reduce bandwidth usage by up to 50%
● Deliver the best quality possible for each user,
based on their browser
Multi-Codec
Streaming
20. Per-Scene Adaptation
Client-Side
● Extracted and process quality data from
content
● Take quality data into account for adaptation
decisions
● Low complexity scenes can be “replaced” by
lower bitrate segments while maintaining the
same perceptual quality
Server-Side
● Extracted quality data during
encoding on a per-chunk base
(using metrics like PSNR, SSIM,
VMAF, or similar)
● Encapsulate and deliver quality
data with the content itself
22. ● Deliver same or better experience while
using less bandwidth (e.g., lower
bandwidth countries or mobile)
● Avoid stalls by predicting high bitrate
scenes
Per-Scene
Adaptation
25. 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 reduces 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.
37. Comparison - Animated Content
● Bitrate Reduction of 28%
● Storage Reduction of 88%
● Quality Improvement 30%
38. Summary
● We want it all -- high quality, low bandwidth!
● Complicating factors get in the way
● Existing solutions only go so far
● New solutions are finally available:
○ Multi-codec
○ New codecs like AV1
○ Per-scene adaptation
○ Per-title adaptation
39. How do I get started?
1) Educate yourself!
2) Research existing strategy and untapped opportunity
3) Contact us to schedule consultation session (sales@bitmovin.com)
4) Work with the team to activate and test selected features on your account
5) Come up with deployment plan
6) Go for it!
Editor's Notes
Mention Google using VP9 for YouTube (only 4K) - https://bitmovin.com/higher-quality-lower-bandwidth-multi-codec-streaming/
Also notice encoding savings due to fewer bitrates in bitrate ladder
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.
Also notice encoding savings due to fewer bitrates in bitrate ladder
Even though top bitrates have a drop in PSNR, it’s still over 45dv for which you will not see an appreciable loss.