Talks

Telepresence in virtual reality can be described as "immersive telebridge," or teleportation to a remote location using a hardware-software complex in the form of a panoramic camera and a server-based distribution and delivery system with a player in the web interface. This solution makes it possible to implement remote assistance services and build products for virtual shopping, presence at events, educational programs with transfer of personal experience, telemedicine, video conferencing and negotiations, construction of paid access to content and other applications.

From the talk you will learn what problems had to be solved to make the streaming of endless HLS streams generative music. Each user's stream is unique and collected on the fly, and interaction with it should not differ from ordinary music listening. The talk will be of interest to those who work with audio streaming.

We will talk about the schedule, sessions, and share the information. Join the broadcast to find out what's on the air soon!

Speaker will talk about how not to get lost in thousand-dimensional spaces, how much you may improve the work of codec by optimal parameterization, and why after having tried to implement this solution by themselves the companies come back to speaker.

In his talk Alexey will discuss what factors affect the stability of voice transmission during videoconferences and speech intelligibility: from sampling rates to the peculiarities of transmitting this data via TCP/IP networks. He will also tell how VK Calls assesses and monitors the audio characteristics to provide users a consistently high level of sound quality.

Nikolai will tell you how WebRTC screen capture and broadcasting is arranged in VK Calls, and why they needed their own implementation for video transmission without loss of quality. The talk touches such technologies as WebRTC, WebCodecs, Media Source Extensions, Wasm. The speaker will also explain how WebM video format is organized and how DataChannel in WebRTC is useful.

Live from the VideoTech 2022 main studio.

Live from the VideoTech 2022 main studio.

Chromium is a big media combine. More than 25 decoders and several demuxers serve to make the tag work flawlessly. The report will talk about the architecture behind it. Why all this variety is missing in Yandex.Browser and other Chromium-based browsers. The speaker will explain how and why you can read media logs in the browser if you create a web with video content.

The success of the service largely depends on the quality of the streams provided. The speaker will talk about system monitoring of broadcasting with QoS and QoE (delivery and playability) parameters tracking. The speaker will describe in which situations the use of objective video metrics is justified, and explain how to interpret the results.

Analysis of the options for implementing streaming with low latency: why and who needs it, and where low-latency comes from. You will learn how live streaming on the Internet is made, why the speaker's team decided to make their own JIT packager and what happened to it.

Now many of us work from home, and even the conferences are held as video calls. One of the technologies which make it possible is called lip sync. It is used to prevent sound and picture from delaying.

Fedor will talk about the theory of A/V synchronization with the example of the task solved by his team when designing a video phone on Android. You will learn how A/V Sync is realized in WebRTC and ExoPlayer, and also what is Multimedia Tunneling. The talk might be interesting for everyone who faced A/V synchronization.

We take stock, remember the bright moments and talk about our plans. Join the broadcast, so you don't miss anything!

We will talk about the schedule, sessions, and share the information. Come to the room or join the broadcast to find out what to expect soon!

The speaker will talk about the difficult battle for fast and high-quality delivery of the first frame in VK Video: why they entered it and what the main difficulties were. What helped the team accelerate and what didn't work.

A look from the bell tower of browsers, frontend and JavaScript. However, all of the principle things would be true for any platform.

Aleksei will share his experience of creating the Teleport Media p2p streaming platform, from 2016 to the present day. He will talk about the architecture of the core (teleport.js script) and backend, as well as key concepts (such as blacklist, in swarm search, timeslot search, hash broadcast, etc.) that allow us to achieve maximum CDN offload. In addition, he will talk about the low-level flaws of the WebRTC datachannel, how they affect the efficiency of file delivery through peering, and what can be done to ensure low latency and HD streaming.

For 13 years, the speaker's company has been selling an archive video recording system for television and video surveillance.

The speaker will talk about the problems that arise when recording video archives. The problems of cloud storage will be highlighted separately.

Daniil will talk about the development and product application of human-centered video analytics technologies, such as multi-camera tracking of people using person re-identification, analysis of their attributes and actions. He will show how a ready-to-use pipeline is assembled from them in a typical way, and where and how it can be effectively used in the end.

Fedor will tell you how the Fastvideo team processes and encodes images. You will learn how to work with them on CUDA and see how fast it is.

As examples, the speaker will look at the projects that Fastvideo has made thanks to this technology.

How to distribute a lot of data without using bgp-anycast technology? The top-level talk will look at the architecture of the CDN layer and the balancing layer when distributing media data. Level of difficulty: for dummies.

An airing from the main studio. Let's talk about what's changed in VK Calls over the past year.

Sooner or later, any service faces the fact that video content costs money. And you have to pay not only for data storage or transcoding capacities, but also for traffic.

In her talk, Olga will look at ways to save the amount of traffic consumed without impairing the user experience.

The report deals with the automation of the establishment of time-limited live broadcasts: from the creation of content cards to the allocation of resources, their management and monitoring. Anastasia will talk about what led to the decision to develop your own product instead of using ready-made options.

The talk focuses on the theoretical basis and implementation details of Forward Error Correction in WebRTC implementation from Google. We will look at what the FEC packet is, how FEC packets are generated, and what points and nuances should be taken into account when implementing and working with FEC.

We will talk about the schedule, sessions, and share the information. Join the broadcast to find out what's on the air soon!

The standard approach in machine learning is to pre-train a neural network that projects the object in question (video, picture, text) into a multidimensional vector space, and then, using these representations, solve other tasks (classification, recommendation, ranking, similarity search).

We will talk about a model for constructing a common embedding space for videos and texts, its training and use for different application tasks.

Live from the VideoTech 2022 main studio.

We take stock, remember the bright moments and talk about our plans. Come to the room or join the broadcast, so you don't miss anything!

Online video streaming services often collect technical statistics from clients to monitor video delivery quality and calculate user happiness metrics. But can they be used in product tasks?

The speaker will tell how his team was able to speed up the calculation of product stats (video viewer count and number of viewers per broadcast) using this data flow and without involving client development.

Live from the VideoTech 2022 main studio.

The talk gives an introduction to SRT: the reasons why it appeared and its features. The talk describes SRT modes and basic concepts, such as latency. Mikhail will share the cases of SRT application and talk about the pitfalls they have stepped on.

Sergey and his colleagues were actively involved in the development of the latest H.266/VVC and MPEG-5/EVC video coding standards. In the report you will learn who and how international standards are developed, how the procedure is organized, and what phases the project goes through from the initial concept to the final specification.

The talk includes an overview of the latest H.266/VVC and MPEG-5/EVC standards. In addition, you will learn about new approaches to image compression based on neural networks, as well as how the new JPEG AI standard is being developed.

Sales of 4K TVs are increasing rapidly. 8K TVs are no longer exotic. The resolution of flagship smartphones has confidently crossed 2K, and tablets have crossed 4K. Such devices will soon require appropriate content. But where to get it? 

Fortunately, Super-Resolution algorithms are being actively developed in parallel. A team led by the speaker has created three public benchmarks of Super-Resolution methods. The talk will consider the problems encountered in answering the question, "Which SR is better?" 

Live from the VideoTech 2022 main studio.

Kirill will talk about the main contradiction of Ultra-low latency using the example of Cloud Gaming. The problem of network losses, principles and mechanisms of forecasting, prevention and recovery after losses, as well as methods of preprocessing, parameterization, and postprocessing of the video stream will be considered.