Pixvana is honored to work with one of the biggest names in the news business, CNN.
CNN is also one of the biggest names in VR content production. Every week the news teams around CNN put out multiple breaking stories in 360 video format.
CNNVR App Takes Viewers Around the World
On March 15, 2018, CNN launched the CNNVR app to takes viewers around the world via news stories streaming on the Oculus Rift. Dynamic 360 content gives CNN the ability to create a truly unique news experience, putting viewers in an active newsroom environment where content changes at the rate of late breaking news. To help power the experience, Pixvana delivered a custom encoding and export system to streamline the on-going publishing of content to the app.
“Together, we are pushing boundaries to realize the full potential of VR to transform the ways news gets delivered,” said Jason Farkas, Vice President at CNN. “Pixvana partnered with our development team to customize the content pipeline and enable our editors in New York to get the highest quality content into the CNNVR app as quickly and as seamlessly as possible.”
Achieving Technical Requirements
High Quality Streaming and Fast Start
CNN drove the requirements for this project, demanding both high quality streaming and fast start for viewers—not an easy technical feat. High quality streaming means delivering 4K content and 4K+ masters with unparalleled visual quality to each VR headset. Fast start is the ability to deliver that high quality content within seconds of the viewer beginning any CNNVR video.
Pixvana built a custom 4K encoding ladder and AVC encoding parameters to optimize the video streams while achieving CNN’s requirement of 15 Mbit network limit. Using custom rate limiting hardware, Pixvana was able to constrain the network and tune the startup behavior across all networks. As a result, CNN will realize cost effective delivery to reach tens of thousands of viewers on the CNNVR app, as well as, reach viewers on more constrained networks outside the US.
Rendering a Complete Content Library
Building the custom encoding ladder required multiple iterations on select pieces of the content library in order to achieve the right bit rate, and ultimately render the complete library–consisting of dozens of 4K masters–in a new format. One challenge was managing the discrepancies in behavior across each piece of unique content that often had slightly different performance delays on startup. To better understand this challenge, imagine the simple, yet tranquil scene of a church with no motion. This 360 video would have a quick start because the data rates for the first minute of viewing the scene are very low and easy to stream. Now, imagine the same startup process, but with a moving shot from a helicopter veering across the New York City skyline. The data rate for this scene can spike quickly and really tax the network, often resulting in performance issues for the viewer. Pixvana solved this challenge for CNN. It’s platform employs up to 50 machines at a time to render the entire library in under 2 hours. This made it easy to respond to ongoing development while meeting CNN’s delivery deadline.
Pixvana’s SPIN Play SDK
For fast start, Pixvana tuned its custom SPIN Play SDK for Windows to do quicker network estimation. The SPIN Play SDK is built on playback technology for Windows 10 to provide MPEG DASH delivery. All views on the CNNVR app will see the best quality that their network— whether Wifi or hard wired—can deliver.
Finally, Pixvana built a custom connector for CNN to move VR media masters to Akamai with correct naming and packaging for streaming from Akamai’s NetStorage system. We are also excited to announce that CNN will be the first Pixvana customer to use a new pipeline API to automatically moves content from AWS to Akamai through a custom encoding and projection system, all under programmatic control.
Check out the CNNVR app to get the latest breaking news in 360 VR from around the world. Then, sign-up for a free trail of Pixvana SPIN Studio to upload and share your own VR content.