October 5, 2026
Comments Off on A Senior WebRTC Engineer’s Take on RTC.ON 2026

For the third year in a row, I had the pleasure of attending the RTC.ON conference where our company WebRTC.ventures was a sponsor. As usual, the panel was stacked with great talks and speakers. The topics ranged from AI, computer vision, voice, audio, telephony, broadcast and transport protocols, including the ever more trending MOQ, which was the most prominent topic at the conference.

You can find the agenda here and the live streams here: day 1, day 2. And below you can find my takeaways from the conference, from my perspective as a Senior WebRTC Engineer at WebRTC.ventures. 

MOQ in Production and in the Spec

Day 1 started with three consecutive talks showcasing production applications using MOQ. 

MOQ Production deployments

The opening speaker was Luke Curley (MOQ co-creator, moq.dev; ex-Discord/Twitch): “On a Boat: Why Pull-Based Media Streaming Beats Push”.  He talked about the benefits of using MOQ over WebRTC and why pull-based streaming is usually better, walking us over a specific use case where the viewer has the power to decide what to ask for, which stream (camera, resolution…) and in which order they need it, and the media will be prioritized according to that.

If you are interested in real production use cases using MOQ you should also check these two talks:

MoQ libraries: imquic and moq-kit 

Besides that, on day 2 we had two speakers that integrated MoQ:

Lorenzo talked about the imquic library (and others) they created to integrate with Janus. It implements both RoQ (RTP over QUIC) and MOQ. It can translate from WebRTC to MOQ and vice versa.

Jakub went over moq-kit, a library they created to publish and subscribe media on iOS and Android natively. He started with the reasons for creating it and the obstacles they encountered, and then went through with the decision making and implementation.

Current debates in the MOQ spec

Going more into the MoQ spec we had 2 speakers that are also closely involved in the development of MoQ:

  • Gwendal Simon (Senior Director of Technologies, Quortex): ”Latest Debates in MOQ: Track Switching and Video Transport”
  • Will Law (Chief Architect, Akamai): ”Making Streaming Faster – MSF, a new format”

Great insights on what is going on with MOQ. The former is specifically about bitrate switching (client vs server). The latter is about the current streaming format discussion.

QUIC and P2P Transport Beyond MoQ 

Going into other transport talks we had:

Jean-Baptiste talked about the importance of real-time remote control and its use cases and presented Kyber, a new project that includes an SDK to stream and control machines using QUIC. He showcased the super low latency with different demos, listed the use cases and went over the features and implementation.

Tim explained why WebRTC supports P2P in the browser, the benefits and the use cases with practical examples. He focused on a Lidar sensor P2P use case, the protocol options he considered and why he chose RTCTransport. He finished with a live demo using Lidar.

On-Device AI Video Effects with WebGPU

We had two interesting talks about on-device AI video effects:

  • Iwo Plaza (Software Engineer, Software Mansion): ”TypeGPU as a foundation for performant on-device media effects”
  • Sam Bhattacharyya (CEO, Katana Video): “Building a Video Effects SDK in the WebGPU and GenAI era”

Both talks were centered around video effects. The former was about using TypeGPU to optimize performance for on-device media effects, with a detailed implementation walkthrough, including a few live demos. While the latter went over a new open-source effects SDK using WebGPU, called Longpipe, which has 10x better performance than MediaPipe, a popular and widely used library. Both are interesting options if you are looking to include video effects into your applications.

AI and Computer Vision for Broadcast and Sports

Same as last year, we had Dan Jenkins (CEO, Nimble Ape): “The LLM in the Gallery: No Director Required”. He built an application for automated broadcast directing. He used an LLM as the director that decides when to change scenes between multiple streams, depending on the inputs, and Smelter as the video composer. It was an interesting proof-of-concept on how to use LLMs to automate broadcast directing an event when there is no budget for a director to switch scenes.

The last speaker of the conference was Piotr Skalski (Lead Open Source Engineer, Roboflow): “Computer Vision, Meet Sports”. He went in depth into computer vision with real-life examples, showing how they built open-source pipelines for sports, in particular for basketball and football. He presented some of the challenges and solutions they found and showed a few demos on how it performs detection, tracking and live stats among others.

Voice, Telephony and Audio at Scale

There were two talks about voice and telephony:

Cezary talked about an open-source project he built called VoiceBlender. It is basically an audio mixer, which allows you to bridge between multiple types of legs (SIP, WebRTC, MoQ, LiveKit, Whatsapp…). He focused on the specific use case of an “AI-powered contact centre” and how VoiceBlender can implement that, with built-in features like answering machine detection and turn-taking. He finished with a demo.

On the other hand, Feliks explained how they built a real-time noise reduction for phone calls using Elixir and Membrane.

There was another talk about audio, but with a different focus:

Stuart explained how they handled transcoding audio files in SoundCloud when you support many upload formats (15+), plus it is on a huge scale, transcoding about half a billion audio files! He walked us through the main challenges they faced, the solutions, the process and the planning. He showed some interesting graphs too.

Mobile Video: Android SVC and Phone Cameras in Broadcast

There was a talk about mobile codecs too:

Pratim went over some issues they faced when using Android hardware VP9 encoders, the debugging process and the findings. Basically, those encoders quietly produce only one spatial layer. He also covered their workaround and gave some practical advice.

This year we also had Daniil Popov (Head of Technology, Cyanview): “Phone cameras in broadcast – from lens to viewers”. He started by explaining color science and how they use it. Then went over the CyanCamera pipeline they created. This resulted in them running an experiment: use phone cameras as a broadcast camera. They were able to demonstrate professional quality in a real festival this year.

Panel, Lightning Talks and WebRTC Live

As a bonus, we had 3 additional programmed sessions:

Discussion panel

At the end of day 1, we had a discussion panel to talk about “Emerging trends in real time multimedia from the practitioner view” with Ben Weekes (Agora.io), Renan Dincer (Cloudflare), Lorenzo Miniero (Meetecho) and Oliver Lietz (Nanoscosmos) as guests, and with Maciej Rys (Software Mansion) as the moderator.

Lightning talks

After lunch on day 2, they organized a session with lightning talks, which consisted of 4-minute talks. We had a bunch of them, mine included: “Real-Time, Fully Local: AI Pipelines with Pipecat”, which was about experimenting with real-time video processing using Pipecat and running AI models locally on a PC.

WebRTC Live episode

Again WebRTC.ventures hosted our monthly WebRTC Live episode from the conference venue. You can find it here WebRTC Live #117: Live from RTC.ON 2026. Hosted by Alberto Gonzalez with some special guests, including: Pratim Mallick, Luke Curley, and Charles Duyk (Principal Engineer on AWS Interactive Video Service (IVS)).

Final Thoughts on RTC.ON 2026

2026’s RTC.ON was an amazing experience, same or even better as the previous years. I had the chance to listen to amazing speakers and interesting talks, meet them and other attendees, talk about all kinds of projects, expand my network and get motivation to build and try new stuff. I am looking forward to next year’s conference!

Recent Blog Posts