The Conference for Video Engineers

Adam Brown

Mux

Running FFmpeg in a docker container is cool, but what happens when you want to leverage hardware encoding technologies such as Intel QuickSync and Nvidia NVENC with docker?

What are the considerations for the host OS and how do these technologies play with existing docker deployment /orchestration tools. Are there performance or feature restrictions, and is it even worth the effort?

Andrew Scherkus

Google

I work on Google's VR team leading our multimedia efforts. For the past two years we've been busy figuring out how best to produce, serve, and playback spatial audio and video. During that time we've realized that there are little-to-no standards for any of this stuff and that it really inhibits the creation and distribution of content.

This talk will cover the basics of spherical representations and the impact they have on visual quality and play back as well as a primer on spatial audio. Considering this is Demuxed, no talk would be complete without diving into our ideas for signalling VR metadata at a container level to enable interoperability between cameras, post-production suites, and players. What else could be more fun than the intersection of sphere-to-plane projections, 3D geometry, and ISOBMFF?!

If you like to be extra prepared, feel free to look at our draft specifications on our GitHub.

Anne Aaron

Netflix

We encode movies and TV shows for millions of viewers worldwide and aim to deliver the best video quality possible no matter what the content, viewing device or bandwidth. This effort hinges on one fundamental premise: that we can accurately and efficiently measure the perceptual quality of a video stream at scale. Simple metrics like PSNR and SSIM do not consistently reflect human perception while large-scale visual subjective testing is practically infeasible.

We developed Video Multimethod Assessment Fusion or VMAF as a method to more accurately measure video quality, and we’ve open-sourced the software. Since this tool applies machine-learning to fuse multiple metrics and features, it is easily extensible by integrating new metrics or adding new subjective training data that can improve the metric. We have used it to evaluate video codecs and encoding strategies. We are in the early stages of using VMAF as part of our automated QC and as a client-side metric to monitor system-wide A/B tests.

Ben Moskowitz

IRC

Overview of rebuilding the PopcornJS / PopcornMaker open source video editor, hooking in to the world's biggest digital library. A large core group of volunteers are coordinating to revitalize the editor to an improved version. Come see what we've been up to or how you can get started, hack on, and volunteer!

Brion Vibber

Wikimedia

HTML5 web video started with black-box video and audio elements, but has expanded with a dizzying array of features from MSE to WebRTC. But what does it look like on the inside? At Wikimedia Foundation, we rebuilt the web video playback stack from the ground up in JavaScript to provide Ogg and WebM compatibility for Safari and IE... This provides a chance to look at the innards of the black box, and how high-level and low-level parts mix. Experience thrills seeing asm.js and WebGL magic, chills at the hacks for streaming URL loading, and prepare to spill your drink as you encounter the remaining low-level limitations to interoperability and performance!

Dan Schultz

Archive.org

Overview of rebuilding the PopcornJS / PopcornMaker open source video editor, hooking in to the world's biggest digital library. A large core group of volunteers are coordinating to revitalize the editor to an improved version. Come see what we've been up to or how you can get started, hack on, and volunteer!

Jai Krishnan

Google

In 2015, Google partnered with Microsoft, Netflix, Amazon, Intel, Mozilla, and Cisco to create the Alliance for Open Media. Our goal is to launch a open and royalty-free codec that is built for the web and is a generational improvement over what is available today, including VP9.

This talk will cover why Google invests in open codecs, the need for an alliance, and where we are a year later with our first codec, AV1.

Nick Chadwick

Mux

WebRTC is most well-known as a browser standard for cross-platform realtime video and audio communications. But its based on industry standards, and industry standards mean RFCs.

In this talk, I'll go through what WebRTC actually is from a protocols perspective, and take a bottom-to-top approach by discussing the RFCs that make up the WebRTC standards.

Starting with ICE, and moving through DTLS, RTP, SRTP, SDP and more, in this talk I'd like to give people a view of what WebRTC actually _is_ , and give people a solid theoretical understanding to get started with this awesome technology.

Nuno Pinge

JW Player

The nuts and bolts of JW Player's implementation of VP9 transcoding and delivery through DASH. Why choose VP9 and DASH even though over 80% of video is delivered with H.264 over HLS today? The choice of technology is being driven by the ever growing needs of our publishers and viewers. We are looking to provide a viable and cost effective alternative to H.264/H.265 over HLS to deliver high quality video for users on constrained bandwidth devices. Here we outline our experiments with quality, encoding speed, bytes delivered and lower time to first frame.

Pooja Madan

JW Player

The nuts and bolts of JW Player's implementation of VP9 transcoding and delivery through DASH. Why choose VP9 and DASH even though over 80% of video is delivered with H.264 over HLS today? The choice of technology is being driven by the ever growing needs of our publishers and viewers. We are looking to provide a viable and cost effective alternative to H.264/H.265 over HLS to deliver high quality video for users on constrained bandwidth devices. Here we outline our experiments with quality, encoding speed, bytes delivered and lower time to first frame.

Steven Robertson

YouTube

YouTube is one of the most widely studied platforms in the space of Quality of Experience. Many of these studies are observational analyses which attempt to translate certain playback events into user retention metrics. The conclusions in these studies are often incorrect.

In this talk, we'll learn how faster playbacks actually lower retention and fatal errors drive more usage. Then we'll learn that that's insane and wrong, and dig in to how those false conclusions were made - and how to avoid making errors that are far more subtle but just as damaging in your own research.

Tracey Jaquith

Archive.org

The nuts and bolts of JW Player's implementation of VP9 transcoding and delivery through DASH. Why choose VP9 and DASH even though over 80% of video is delivered with H.264 over HLS today? The choice of technology is being driven by the ever growing needs of our publishers and viewers. We are looking to provide a viable and cost effective alternative to H.264/H.265 over HLS to deliver high quality video for users on constrained bandwidth devices. Here we outline our experiments with quality, encoding speed, bytes delivered and lower time to first frame.

David LaPalomento

Brightcove

Media Source Extensions promise some amazing new capabilities for web video. Once you take that shiny new API out of the garage though, you'll discover a number of delightful quirks that can make it quite difficult to be successful with. This talk will go over a bunch of rough edges in the spec and implementations, discuss popular workarounds, and propose some practical improvements that could make MSE easier to work with.