Monthly Newsletter
WHAT'S RTC.ON NEWSLETTER ABOUT?
Each month, our team will get together and give you a selection of multimedia dev content.
Grab yourself a coffee (or any drink of preference), sit back & enjoy the next edition of RTC.ON Newsletter – we hope you'll like it!
MULTIMEDIA MUST-READS
Wojtek Barczyński
SOFTWARE ENGINEER @ SOFTWARE MANSION
How to lie about latency
Hardly a week goes by without people in the real-time video dev communities and on social media boasting about their jaw-dropping latency result. More often than not those results are cherry-picked and are far from real glass-to-glass results. An article from Open Broadcast Systems explains what the catch is. It covers the most popular lies in those latency measurements.
28 Days Later (my QUIC journey)
Lorenzo Miniero explains how he went from reading QUIC RFCs to integrating WebTransport to Janus. The personal, journey-style format of the blog gently introduces and explains complex topics one at a time. There’s also a great follow-up blog called RoQ ‘n’ roll!). Moreover, I highly recommend checking Lorenzo’s awesome talk about Bandwidth Estimation in the Janus WebRTC Server from last year’s RTC.ON.
Writing an MP4 Muxer for Fun and Profit (Except there is no profit, only pain)
This blog explains the basics of MP4 structure, highlights differences between regular and fragmented MP4s, and presents an interesting “hybrid” approach that will be shipped soon to OBS and FFmpeg. It leaves readers with a good high-level overview of how MP4s are structured, without scrambling their brains with complex details :)
Enhancing Netflix Reliability with Service-Level Prioritized Load SheddingAutor: Netflix TechBlog
An article from Netflix describes their new approach to service-grained load shedding. It explains how they started prioritizing user-initiated requests over prefetch requests for PlayAPI and generalized it for other, not only video-related services.
Bringing native AI to your mobile apps with ExecuTorch
Jakub and Wojtek from Software Mansion’s AI team are diving into ExecuTorch - a novel PyTorch-based framework for local deployments on devices like mobile phones and microcontrollers. As an example, they present how you can bring LLM into an image editor app.
Google Chrome Secretly Shares Your Device Data - With Google
The article reports on Luca Casonato’s) claim of Google Chrome granting full access to system/tab CPU, GPU, and memory usage only to *.google.com sites. Other sources provided a method to replicate and confirm this claim. It’s unclear if meet.google.com/ use this feature.
RTC.ON 2024 TALKS DIVE-IN
Karolina Kulig
MARKETING MANAGER @ SOFTWARE MANSION
Everything you wanted to know about VLMs but were afraid to ask
Piotr is a Software Engineer working at Roboflow, a company creating SAAS products making building with computer vision easy. During RTC.ON, he’ll dive into Vision Language Models (VLMs) – one of the fastest-growing niches in Computer Vision.
Take a masterclass from the computer vision Expert – learn all the basics, see real-life examples and ask your questions during a short Q&A session after the talk!
Stop Fighting Hydra: Replacing Headless Chromium
Wojtek works at Software Mansion, where he co-creates the Live Compositor – a media server dedicated to real-time video & audio composition. During RTC.ON, he’ll discuss headless chromium, asking a question of whether it’s the best solution for real-time video mixing (spoiler: it isn’t).
WebRTC and QUIC: how hard can it be?
In his RTC.ON talk, Lorenzo Miniero, the author of Janus WebRTC Server & organizer of JanusCon, will delve into using QUIC for real-time multimedia applications. Lorenzo did the researching & testing job for you – join him for his presentation and see what the buzz around QUIC is all about & if/how hard it would be to get it and WebRTC to like each other!
Building Your Own RAG Model: A Step-by-Step Guide to Interact with Your Documents
Paula Oses, an AI Engineer at IAG, will explore the use of Retrieval Augmented Generation (RAG) with Large Language Models (LLM) to interact with your own documents. The purpose of the talk is for the listeners to gain a comprehensive understanding of how to develop their own RAG model & unlock all the potential for dynamic interaction with the documents. After the talk, you’ll have the opportunity to ask Paula your questions regarding RAG.
Intrigued by the talks? Click the button below and see all the announced RTC.ON 2024 speakers. And yes, there is even more to come!
WHAT'S NEW IN SOFTWARE MANSION MULTIMEDIA TEAMS
Maciej Rys
SOFTWARE ENGINEERING LEAD @ SOFTWARE MANSION
Live Compositor
The multimedia toolkit built at Software Mansion has an exciting new addition: Live Compositor. It is a media server written in Rust that provides capabilities for real-time, low-latency, programmable video and audio mixing. Live Compositor is a perfect tool for a variety of use cases, including broadcasting, video conferencing, and live streaming. The latest showcase video featured below presents the main features of the product and explains best the idea behind it. Curious what’s coming next? Join us at our Discord channel.
ExVision
ExVision is our attempt to deliver a library that is dead simple to use for AI vision-related models in the Elixir ecosystem. With the recent 0.3.1 release we extended the range of supported models in the space of classification, object detection, semantic segmentation, instance segmentation, and keypoint detection. If you have any feedback or feature requests do not hesitate to reach out here.
Fishjam Cloud
It’s already been a couple of weeks since we announced the launch of Fishjam Cloud - our platform that aims to make building video streaming apps as simple as possible. During this time, we have had many conversations with people interested in early access. We have learned a ton and are even more excited about the product we are about to launch. Nevertheless, we believe there is no such thing as too much feedback and customer insight. If you would like to discuss your use case with us and participate in early access, please sign up!
Elixir WebRTC
Batteries included WebRTC implementation for the Elixir ecosystem coming in September – stay tuned!
Thanks for making it this far!
Happy streaming :)