Siperb’s Go-Based Media Bridge Lets WebRTC Softphones Work With an On-Premises PBX (Siperb)

Read it twice and the news gets smaller each time. Siperb built a Go media bridge that runs inside its own desktop client: it terminates the WebRTC session locally. That is a WebRTC-aware SBC or a variation of it, and the release says as much one paragraph after explaining that the whole point was to avoid shipping one. The SBC is still in the call path, it just ships inside the softphone now: no appliance in the building, no cloud hop, and one more piece of media plumbing to get right on every customer’s machine. Obviously, they didn’t forget to mention it was written in two days with an AI coding assistant, against a spec modeled on RTP proxy design. But what is it they are really trying to announce here?

📰 This Week in WebRTC

• From Bounding Boxes to Motion (Loic Denoue) - tracking mouse cursor in a browser screen sharing session

• Shipping Video Calling on Facebook: the agony and ecstasy of building the first version of Facebook video calling (Philip Su) - How time flies when you’re having fun. The days before WebRTC seem so... old

• How to Build an AI Voice Agent with Python and WebRTC (Stream) - a cookbook for STT-LLM-TTS loop, turn detection, barge-in. All the things OpenAI is now changing

• Control your video calls with gestures using free AI from Google (OpenVidu) - gestures are so 2024. I wonder who asked for it from OpenVidu

📣 Sponsors

Meet Sipfront live on 26th August
Discover the all-in-one voice and WebRTC testing platform trusted by NFON, Parloa, and sipgate. Join Sipfront for a live demo on 26 August.

vCon: The PDF for Your Conversations
An open standard for transcripts, metadata, and consent. Learn vCon.

✍️ This Week on BlogGeek.me

OpenAI just changed how Voice AI is built - OpenAI deleted the turn detector rather than improving it. GPT-Live-1 splits voice from thinking into two tracks with a swappable frontier model, and takes WebRTC call setup from six round trips down to one. A brave new architecture

⭐ Tsahi’s Pick

Built with WebRTC, with Justin Uberti (OpenAI) (Built with WebRTC)

This week, I am picking Justin. I sat down with Justin Uberti - he created WebRTC at Google, founded Ultravox, and now runs real-time infrastructure at OpenAI - the same week OpenAI published its GPT-Live architecture post. 44 minutes of your time to understand how OpenAI fundamentally changed Voice AI architectures moving forward, and why WebRTC still needs working on - especially when you are talking to machines instead of other mere mortals.

📰 This Week in WebRTC

• Voice AI evals score the pipeline, not the network (Sponsored by rtcStats) - evals score the STT-LLM-TTS pipeline, but two transport failures leave no mark in a transcript, and full duplex removed the slack that used to hide them

• Tor launches Snowflake app to bypass censorship blockages (Cybernews) - Snowflake now ships as its own Android app. It uses WebRTC as camouflage, though it is uncertain if that’s also part of this app

• Running STUNner in Production on Kubernetes (WebRTC.Ventures) - a step by step guide for setting up STUNner on Kubernetes

• Top 5 Things to Check When Selecting a Video Call API (Whereby) - a nice starting point from a vendor offering his own API. There’s more things to check and KPIs to review, so use this as a starting point

• How to Implement WebRTC Adaptive Bitrate Control: Tutorial (Red5) - my only question is why not let WebRTC do that dynamically on its own instead of in the application

📣 Sponsor

Read about the latest WebRTC.ventures Story of Success.
Self-hosted Jitsi on Kubernetes, built for a telehealth platform to replace meeting links with real ringing and answer, on infrastructure the client now runs themselves.

📢 Releases

• GStreamer 1.28.6 - the final 1.28 release adds H.266 MP4 muxing and FFmpeg 9.0 support, along with DTLS and WebRTC adjustments for OpenSSL 4.0.0

• Digital Samba 644 - local mute, .mov uploads, and call quality reporting

⭐ Tsahi’s Pick

How we built a realtime system for responsive voice AI in six months (OpenAI)

OpenAI changed the game in Voice AI. It took the turn detector out of the audio path entirely and let the model decide when to speak. They cut WebRTC call setup from six round trips to one with WARP. Part of the changes made were architectural. This is going to affect all Voice AI vendors and frameworks - commercial and open source. This is going to be interesting to see how it develops.

📰 This Week in WebRTC

• Streaming the IETF MoQ sessions with MoQ… kinda! (Meetecho) - Meetecho streamed the IETF 126 MoQ sessions over MoQ itself, and writes up honestly where it still needed help (hint: it is everything that WebRTC gives for free)

• An API for MoQ: provision your own isolated relays (Cloudflare) - Cloudflare is more vocal (and faster?) in its adoption of MoQ versus its adoption of WebRTC

• How to Use chrome://webrtc-internals to Debug Calls (Digital Samba) - the 101 guide of the week. This time, on webrtc-internals

• Professional TURN Servers for WebRTC (Chris Ahrweiler) - because what we were all missing was another managed TURN service

📣 Sponsor

vCon: The PDF for Your Conversations
An open standard for transcripts, metadata, and consent. Learn vCon.

📢 Releases

• Chrome 151 release notes (Chrome) - data channel setup now carries the SCTP handshake inside the SDP offer, cutting up to two round trips off establishment. Fits as part of the OpenAI story

• Native Video Reference App 1.3 - Vonage’s iOS and Android reference app adds background replacement, an extensible bottom bar, and new call controls

• coturn 4.16.0 - a new release of the TURN/STUN server, landing the same week Fedora pushed 4.15.0 out as a security update

• rtcStats v1.9 - self-hosted enrichment through the rtcstats-enrichment server

✍️ This Week on BlogGeek.me

WebRTC won’t fix your call quality. Observability will - WebRTC’s call quality is no better than any other VoIP protocol, and a discovery call reminded me why teams keep expecting otherwise. What separates the ones who fix quality from the ones who keep guessing is whether they can see what is actually happening

⭐ Tsahi’s Pick

Making 1080p screen shares look sharper (Hopp)

Hopp is focusing on remote pair programming, and as such screen sharing is important. Here’s the thing. Screen sharing goes at 4K and above most of the time. If the source is 1080p and you upscale it to 4K, the result isn’t sharp enough. Here, adding a sharpening filter at render time got a visibly crisper image for free. I like this one because it is a reminder that perceived quality in WebRTC is not only a bitrate problem - and upscaling can be improved upon.

📰 This Week in WebRTC

• Recovering from service disruptions with the Vonage Video API (Vonage) - Vonage explains their session disruption callbacks, which is an interesting concept

• Agentic workflows for WebRTC and real-time development (WebRTC.ventures) - WebRTC.ventures sharing their experience with AI-assisted development in WebRTC

• How DynDevice built virtual classrooms into its LMS with OpenVidu (OpenVidu) - a case study of DynDevice LMS choosing OpenVidu to replace video meeting SaaS platforms

• Network traffic engineering: WebRTC vs ST 2110 (The Broadcast Bridge) - the failure here for me is the not understanding that managed networks are becoming rarer as time goes by

• Your session data shouldn’t leave your VPC by default (rtcStats) - rtcstats-server keeps WebRTC session data inside your own infrastructure; forwarding anything out to rtcstats.com is opt-in only

📣 Sponsor

WebRTC Live is a monthly webinar series featuring real-time industry guests, hosted by WebRTC.ventures.
Join us for the August 12 episode, “Building Robust Voice AI Pipelines for Every Use Case” with Varun Singh of Daily.co / Pipecat.ai.

📢 Releases

• libaom v3.14 “Pinkie Pie” - AOMedia’s AV1 reference encoder/decoder ships eight months of encoding and performance improvements since v3.13

⭐ Tsahi’s Pick

Accelerating automotive innovation with C4A-metal and Panasonic Automotive vSkipGen (Google Cloud)

Most people still read WebRTC as a video calling technology. Panasonic is using it as the last mile into a cloud-hosted digital twin of a vehicle cockpit: the full Android Automotive stack runs on bare metal in Google Cloud, the GPU renders the HMI remotely, and a standard browser becomes the display. For me? That’s just cloud gaming for non-gamers. Get a behemoth of a machine in the cloud to render your “game” or vehicle cockpit - and then stream live to a browser to interact with it. I wonder when the MoQ people will start drooling on that use case as well...

📰 This Week in WebRTC

• The real cost of building WebRTC video (Digital Samba) - the cheap prototype is only the start; this maps the engineering, infrastructure, QA, compliance, and maintenance costs that follow

• Managing permissions in video conferencing apps: 3 access models (OpenVidu) - three ways to model conferencing permissions, and the trade-offs behind each

• MoQ in the Fast Lane (Luke Curley) - Luke raves about MoQ (again) and how WebRTC botched it on things you can do with WebRTC. Inventing stuff is always more exciting

• Reverse-remote OB trucks use WebRTC for multiviewer monitoring (SVG Europe) - broadcast trucks. Guess what they use for monitoring multiviewer streams?

• You don’t need another dashboard for WebRTC observability (rtcStats) - Tsahi on why WebRTC observability should answer the debugging question instead of adding another dashboard to watch

📣 Sponsor

vCon: The PDF for Your Conversations
An open standard for transcripts, metadata, and consent. Learn vCon.

📢 Releases

• LiveKit Unity SDK 2.0.0 - takes the Unity SDK out of Developer Preview for production voice, video, and AI experiences in games

• LiveKit Video Frame Metadata - sync media with metadata

✍️ This Week on BlogGeek.me

The vendor you’re about to bet on can’t see the market. Neither can you - Every vendor sees the market through its own customers, lost deals, and internal conversations. The dangerous part is everything outside that view: buyers who never shortlisted you, competitors you never noticed, and AI systems making decisions with incomplete context. That blind spot is why I rebuilt the 2026 Video API report from the outside in.

⭐ Tsahi’s Pick

ONVIF releases Profile V draft for cloud video without vendor lock-in (SDM Magazine)

WebRTC is about 15 years old now, and it keeps getting pulled into standards bodies that have been around far longer. We saw it with Matter 5.0 in the smart-home world, and now ONVIF - the surveillance and physical-security standards body - has made WebRTC the mandatory transport in their Profile V draft for cloud video. WebRTC is not winning because it’s the obvious choice for video calling - it’s winning because it’s a well-understood, browser-native transport that established standards bodies trust enough to build on.

📰 This Week in WebRTC

• Google Summer of Code 2026 - meet this year’s Jitsi projects (Jitsi) - five GSoC contributors are working inside the Jitsi stack this summer. An interesting selection

• Mediasoup DataChannels: when replacing WebSockets actually makes sense (Jesús Leganés-Combarro ‘piranna’) - a detailed walkthrough of when mediasoup data channels actually beat WebSockets - as well as when you shouldn’t use it

• We deployed the same video platform on five clouds and timed it: 5 minutes to 20, and the slow ones… (OpenVidu) - when you let Claude (or another AI?) run the deployment and then write the post for it

• Profile V – Frequently Asked Questions (ONVIF) - ONVIF’s new draft spec standardizes cloud video transport with WebRTC as a primary, mandatory path. They also define H.264 as the video codec

• WordPress Playground: remote access over WebRTC (WordPress Developer Blog) - Playground now lets you open your local dev environment from another device over a direct browser-to-browser WebRTC connection

• Call quality is invisible until a customer leaves (No Jitter) - Tsahi on how contact centers lose sight of WebRTC quality because they don’t own or control the networks and devices of the agents

📣 Sponsor

WebRTC Live is a monthly webinar series featuring industry guests (WebRTC.ventures)

Join us for the July 22 episode, “Using Voice AI and Avatars to Scale Live Training” with guest Andrés Rincon.

📢 Releases

• GStreamer 1.28.5 - some WebRTC improvements and refinements in this release

• Red5 Pro v15.6.0 - adds AV1 codec support, though unknown if it also works in the WebRTC protocol for Red5

• OpenVidu 3.8.0 - mediasoup is now a fully supported RTC engine with all previous limitations resolved; OpenVidu can now leverage it end-to-end

• rtcStats v1.8: dark & light themes and richer overview statistics - dark mode, a redesigned overview with min/max, P95, and standard deviation; streams timelines now anchor to call start

⭐ Tsahi’s Pick

Why Every Robot Teleoperation Team Eventually Outgrows WebRTC (RoboticsTomorrow)

Every few months someone writes the “we outgrew WebRTC” post, and this is a good one because it’s specific about where the wall is. The funny thing is that most just parrot each other, on where the default behavior of WebRTC fails them in browsers. They forget the fully own and control the sender side and can still tune the receiver end. Instead of grokking WebRTC, they go inventing it themselves. Time and again.

📰 This Week in WebRTC

• Scaling Janus WebRTC Server: Building a Media Resource Broker (WebRTC.ventures) - one of the first times I see horizontal scaling of SFUs explained in such detail

• Building a Video AI Agent with Vonage Video Connector SDK and Pipecat Transport (WebRTC.ventures) - Vonage added Pipecat support, and WebRTC.ventures took it for a spin

• SVC vs Simulcast in WebRTC | Complete Comparison 2026 (Digital Samba) - a look at the differences between SVC and simulcast. A nice read

• Vector Vibing to speed up Opus encode by 20% (WebRTC Hacks) - Philipp Hancke continues to pick and choose areas to fix and optimize. This time, taking Opus encoding down 20% CPU on desktop

• How OpenAI Delivers Low-Latency Voice AI for 900M Users (ByteByteGo) - a regurgitation of the original OpenAI post, along with a few limitations cited near the end of it

📣 Sponsor

vCon: The PDF for Your Conversations

An open standard for transcripts, metadata, and consent. Learn vCon.

📢 Releases

• Chrome 150: WebRTC Diagnostic Logging API - apps can now capture WebRTC diagnostic logs programmatically instead of leaning on webrtc-internals by hand. The better approach is to use rtcStats, but I am biased about it...

✍️ This Week on BlogGeek.me

What getting rtcStats seen taught me about the whole WebRTC ecosystem

I spent months trying to get rtcStats noticed, and it turned into an accidental tour of how the WebRTC ecosystem actually pays attention: who links, who ignores you, and what it takes to register at all. Here’s what that taught me about the space I’ve been covering for years.

⭐ Tsahi’s Pick

Join Google Meet calls from Safari on iOS devices (Google)

Google Meet now runs in Safari on iPhone and iPad, no app required. For years Safari was the browser you warned users about when building on WebRTC, and Google kept Meet in its iOS app for a reason. If Google is now comfortable putting Meet into mobile Safari, Apple’s WebRTC support has quietly become good enough for one of the most demanding real-time apps out there. The question is whether the rest of the vendors follow and start trusting iOS Safari.

📰 This Week in WebRTC

• How to debug a webrtc-internals dump with AI (rtcStats) - the AI-assisted counterpart to reading the dump by hand: upload one and get the what-went-wrong in under 30 seconds

• POC time: a push-to-talk demo with MoQ (Meetecho) - Lorenzo keeps poking at MoQ, this time a working push-to-talk demo bridged to WebRTC and back

• 5 Reasons MoQ Streaming Is a Great Fit for Surveillance and Monitoring (Red5) - This reads future looking, based on the fact that implementations of MoQ don’t interoperate when not on the same draft yet

• Berlin’s Almetra raises €16.3M to turn factory video into live data (The Next Web) - One wonders if live here means WebRTC or something else

• AV1 in 2026: Why the ‘Next-Gen Codec’ Still Isn’t Dominant (Trembit) - I just love reading AI generated content and losing interest when the whole premise falls apart - 90% of WebRTC sessions use VP8 - that’s when Google Meet defaults to VP9 and uses AV1 - and are the biggest WebRTC player in town...

📣 Sponsor

Beyond WebRTC Projects: Build Your Complete Engineering Team (WebRTC.ventures)

You’ve reached a pivotal moment. Your WebRTC application is successful, user adoption is growing, and you need to expand development capacity significantly.

✍️ This Week on BlogGeek.me

• How to Read a webrtc-internals Dump, Section by Section - my walkthrough of a webrtc-internals dump, section by section, so a bad call’s recorded data actually tells you what went wrong

⭐ Tsahi’s Pick

Adopting AV1 for Real-Time Communication (RTC) at Scale (Meta)

Meta walked through what it actually took to get AV1 running in real-time calls across most mobile devices. A lot of effort is spent in introducing a new video codec, and this is on the extreme end of things at Meta’s size. This includes CPU, memory, binary size, adaptivity, presets, optimizations and more. An interesting read on what it takes to adopt AV1.

📰 This Week in WebRTC

• Can MoQ and WebRTC be friends? (Meetecho) - Lorenzo Miniero on where MoQ and WebRTC actually meet

• Open Live: Building a Software Broadcast Mixer on Commodity GPU Hardware (Eyevinn) - Eyevinn moves the broadcast switcher to software on off-the-shelf GPUs

• AI Video Agents With Vonage, Pipecat and AgentCore (Vonage) - After the official announcement, Vonage leans down on Pipecat with a howto post

• 5 conversational AI app ideas you can build this weekend (OpenVidu) - “Claude - write a post on what AI apps can users build with AI”

• rtcStats vs webrtc-internals in 2026: a real comparison (rtcStats) - when the built-in tool is enough and when you need production monitoring

• What nobody tells you about running WebRTC in production - lessons from 15 countries in 9 months (Reddit) - well... you could read about WebRTC... or take a training course and you won’t be able to complain about nobody telling you things ;-)

• Stream Edge AI Vision to Any Browser via WebRTC (Hackster.io) - obviously, it needs RTSP-WebRTC to get the stream from the camera to the browser

• How forward and backward secrecy actually work in a multi-party video call (Digital Samba) - a look at the challenges and decisions needed by E2EE implementations

• The VLC developer betting on millions of robots needs millisecond precision (WebProNews) - Kyber is betting on QUIC and not WebRTC for low latency streaming

📣 Sponsor

vCon: The PDF for Your Conversations
An open standard for transcripts, metadata, and consent. Learn vCon.

📢 Releases

• Wowza Streaming Engine 4.11 - reimplemented WebRTC engine with selectable v1/v2 modes (release notes). Is Wowza finally serious about WebRTC?

• rtcStats v1.7: Client Insights, dedicated API host, smoother onboarding - Devices page now centers on client insights, and the public REST/MCP API moves to its own host (api.rtcstats.com) with an observations catalog

✍️ This Week on BlogGeek.me

• WebRTC for Voice AI: How the transport layer works in 2026 (and where it’s headed) - my breakdown of WebRTC as the transport for voice AI, how it stacks up against WebSocket and WebTransport, and where to optimize past the defaults

⭐ Tsahi’s Pick

How We Moved Discord Voice to the Edge (Discord)

Discord moved 80%+ of its voice and video onto Cloudflare’s edge, going from around 30 locations to 300+ cities. The story is interesting and sheds a light on large scale operations and what happens when you migrate them. The two interesting tidbits there for me? 25k voice hosts; and the fact that their SFU is written in Rust. From in-between the lines, I’d suggest to them as a next step do decouple meetings from SFUs and use cascading for their topology.

📰 This Week in WebRTC

• Voice AI Security: Building Realtime Voice Agents with WebRTC, LiveKit, and Sensitive Data Guardrails (webrtc.ventures) - How to keep sensitive data from leaking once your voice agent is wired up to an LLM

• Bridging RTSP to WebRTC on iOS (Medium) - What I failed to understand is the reasoning behind connecting an external RTSP feed via an iPhone to some other user over WebRTC. Doing it on a Linux server would have made total sense here

• Live-streaming CommCon 2026 with MOQ: a guide and a cautionary tale (Nimble Ape) - MOQ is so great - it can do everything you want! Probably because you need to write everything yourself

• OpenAI WebRTC Audio Session, Now With Document Context (Simon Willison) - if you can’t wait to use their latest voice model... Willison got you covered

• How WebRTC Quietly Rebuilt Real-Time Video on the Web (GISuser) - this is interesting from the perspective of geographic data and the implications to it. The rest is WebRTC basics

📣 Sponsor

WebRTC Live is a monthly webinar series featuring industry guests, hosted by WebRTC.ventures
Join us for the June 24 episode, “An Insider’s Look at OpenAI’s Rearchitected WebRTC Stack” with guest Yi Zhang from OpenAI.

📢 Releases

• WebRTC in Chromium M150 - Release notes for the next Chrome milestone landed on discuss-webrtc (we got M149 on the same day as well)

• Vonage React Native SDK 2.33.0 - Adds publisher degradation preference, so you decide whether resolution or frame rate drops first when bandwidth gets tight

• mediasoup 3.20.3 - Ships an important SCTP fix

• GStreamer 1.28.4 - Resolves some regression issues with WebRTC H.264 use

⭐ Tsahi’s Pick

Microsoft Teams Android Rooms Get SIP Cross-Platform Upgrade (VoIP Review)

Two Microsoft Teams products moved away from WebRTC this week - Android Rooms drops WebRTC Direct Guest Join for SIP interop, and Teams on AVD and Windows 365 swaps out the legacy WebRTC optimization layer for a new media engine. It saddens me to see that, and if feels as if Microsoft hasn’t done enough to go all in with WebRTC at any point - or that’s the story I tell myself here. We’ll see how other vendors play it out in the next 12 months.

📰 This Week in WebRTC

• What’s New in Hosted Agents in Foundry Agent Service (Microsoft Foundry Blog) - Teams moves away from WebRTC while Agents move towards it. Go figure...

• Connectivity Resilience and Security in WebRTC Deployments: Key Considerations on TURN (OpenVidu) - Where OpenVidu boasts on how they configure and manage TURN servers to remove your headaches

• RTMP vs. SRT vs. WebRTC: Choosing the Right Live Streaming Ingest Protocol (Mux) - Always interesting to see how vendors who have little to do with WebRTC in their solution frame it

• Open Source WebRTC Media Servers: Choosing the Right One for Your Use Case (webrtc.ventures) - A nice overview of open source WebRTC media servers available today

• Chrome Patches 429 Vulnerabilities (Cybersecurity News) - Multiple use-after-free bugs in WebRTC among the 429 CVEs fixed in Chrome 149. The power of AI...

• Safari Technology Preview 245 (WebKit Blog) - The crumbs we get from Apple as WebRTC fixes. This time, 2 related to video

• An app that lets anyone control a robot from their phone, no coding required (Digital Trends) - Robotic teleoperation by way of WebRTC

📣 Sponsor

From dump to root cause: 60 seconds (rtcstats.com)
Drop your WebRTC dump in rtcStats. Root cause in seconds.

✍️ This Week on BlogGeek.me

• The Real Cost of Real-Time: What Your WebRTC Bill Actually Looks Like - What to look for when trying to calculate price estimates with Video API vendors - and where the gotchas are hidden

⭐ Tsahi’s Pick

Vibe-coding WebRTC

Read past the clickbait title. Fippo is back to landing real changes in libWebRTC, Chromium, Firefox, and pion, including a getStats throttling fix he filed in 2022 and the cryptex RFC 9335 work. What changed is not that AI writes WebRTC. What changed is that an engineer with deep context can now operate across codebases he does not write daily, including alien ones like Go. The bottleneck used to be the time cost of crossing into unfamiliar territory. That bottleneck just got smaller, and it shows in what ships.

📰 This Week in WebRTC

• The Architecture of Scale: How to Scale Video Conferencing - Engineering walkthrough from single SFU box up to multi-region cluster, with the decision points along the way

• What It Takes to Ship a Production AI Avatar System - The hidden line items in an AI avatar build: latency budget, sync, fallback paths. Useful if you are scoping one

• Server-Side WebRTC Noise Reduction with Pion, FFmpeg, and RNN Models - Practical Go pipeline: receive media via Pion, pipe to FFmpeg, run an RNN denoiser server side. The fallback when you cannot push DSP to the client

• Recording True Stereo Audio Using getUserMedia - Walks through the constraints, gotchas, and browser behavior for capturing real stereo from getUserMedia

• SecureConf Announces Brandable Web Conferencing Platform - Yet another white-label conferencing entrant. Worth a glance if you track the long tail of build-vs-buy contenders

• What Is Media over QUIC (MoQ) and Why It Matters for Real-Time Streaming - Because we didn’t share any MoQ vs WebRTC post for about 2 weeks now...

• EDuke32 in the browser, multiplayer finally works - Play Duke Nukem 3D multiplayer in the browser using WebRTC. What more can one ask for?

📣 Sponsor

WebRTC Live is a monthly webinar series featuring industry guests, hosted by WebRTC.ventures.
Join us for the June 24 episode, “An Insider’s Look at OpenAI’s Rearchitected WebRTC Stack” with guest Yi Zhang from OpenAI.

📢 Releases

• Vonage: Video Connector SDK and Pipecat Transport for AI Video Apps - New SDK plus a Pipecat transport for plugging AI video pipelines into Vonage Video

• Digital Samba release 639: Content Preview, Unmute Feedback, Poll and Quiz Limits - Content preview before sharing, signal feedback on unmute, and explicit caps on polls and quizzes per room

• rtcStats v1.6.0: Analysis Page Revamp, Rich API Filters, PC Security - Reworked analysis page, richer API filters, and PeerConnection-level security hardening

⭐ Tsahi’s Pick

Daily Recording Upgrades: Raw-Tracks & New Compositor

Recording used to be simple in WebRTC. You did it on the client side, with MediaRecorder. Or used the server for that, either by combining media streams (=mixing) or saving as separate files. Daily is taking it to the extreme, offering code for doing the compositing they do in realtime for those doing it after saving separate files. I had that headache once with a client of mine - having this done for you is great. Kudos for adding this!

📰 This Week in WebRTC

• Sweden goes live with Matrix-based federation! - Two Swedish public sector agencies federate using Matrix. That’s a win for both Element and Matrix

• Fixing WebRTC data-channels head-of-line blocking with RFC-8260 - Pion implements SCTP message interleaving so small messages don’t wait behind large fragment transfers on data channels

• What is dav1d and Why it Matters - AOMedia on their open-source AV1 decoder (in classic heartless-llm-content style)

• Why Live Video Compositing Still Requires a DevOps Team - Daily plods along about recording and Fishjam explains the basic challenges

• Palabra.ai hits $1M ARR, grows 17x in six months - Real-time AI voice translator built on WebRTC

• Lawo showcases Edge One at InfoComm 2026 - WebRTC-based broadcast production platform demo at InfoComm

• Building AI Voice Assistants with Google Gemini 3.1 Live and Angular - Streaming Gemini voice with WebSocket and adding video through WebRTC (why not voice as well?)

• Scaling Up is easy, the challenge is Scaling Down - OpenVidu on the scale-in problem

📢 Releases

• New Video API 2.34: Network Insights and Expanded Browser Support - Vonage adds real-time network condition reporting and extends background blur/replace to Safari and Firefox

• New Video API 2.34: Network Insights and Expanded Browser Support - Vonage adds real-time network condition reporting and extends background blur/replace to Safari and Firefox

• What’s New in Red5 Pro v15.5.1 - Red5 adds RTX support to WHIP

📣 Sponsor

rtcStats - From webrtc dump to root cause. 60 seconds.

Stop staring at webrtc-internals dumps. Upload a call to rtcStats and get root cause analysis in under a minute.

✍️ This Week on BlogGeek.me

• Chrome moves to a 2 week release cycle. Where are you with your WebRTC app? - Chrome’s shift from 4-week to 2-week releases means less time to catch WebRTC breakage before it ships

📖 Reading

Every Voice and Video Call on Discord Is Now End-to-End Encrypted (Discord)
100% E2EE support across all its client apps.

Tesla reveals two Robotaxi crashes involving teleoperators (TechCrunch)
Tesla has teleoperation function. I wonder… is it WebRTC based?

⚙️ Technical

Voice Agent testing by the experts at WebRTC.ventures (sponsored)
Bringing ten years of WebRTC engineering and production-grade QA to the rapidly evolving field of AI voice agent testing.

Build real-time voice streaming applications with Amazon Nova Sonic and WebRTC (AWS Blogs)
AWS shows off connecting their Nova Sonic speech-to-speech model to Amazon Kinesis.

Real-time voice agents with Stream Vision Agents and Amazon Nova 2 Sonic (AWS Blogs)
The same AWS connects Nova Sonic to Stream’s Vision Agents framework.

Tencent Cloud and Stream Partner to Accelerate the Development of Real-Time Multimodal AI Agents (Tencent RTC)
Tencent picks Vision Agents as its AI framework via partnership with Stream.

💼 Use Cases and Customer Wins

Rethinking BYOD security: protecting data without trusting devices (Ubuntu blog)
Zero-trust BYOD security using WebRTC streaming.

How AI is rewriting Customer Support (WebRTC For Developers)
Olivier spills the beans on our immediate roadmap at rtcstats.

📢 Releases

Fishjam Release Notes: 0.27.0 (Fishjam)
Changes in how credentials are managed.

What’s new in STUNner v1.2.0 (L7mp Technologies)
Backend work and optimizations.

🤓 From our own posts

• MOQ is lacking a compelling adoption reason (BlogGeek.me)

That’s all for this week. See you next week!

📖 Reading

Pexip and Wire join forces to advance sovereign European communications (Press release)
An interesting partnership between two Europeran vendors - video and messaging.

⚙️ Technical

From webrtc dump to root cause in 60 seconds (sponsored)
Drop a webrtc-internals dump into rtcStats. Get an experience score and automated findings in under a minute.

Peerix: WebRTC development made simple (Meefik’s Blog)
A new P2P abstraction framework. As always, question is if this is going to be maintained in the long term.

How the networks of your clients affect their user experience and your server infrastructure costs in a WebRTC platform (OpenVidu)
OpenVidu with a nice explainer of NAT traversal from a media server perspective.

💼 Use Cases and Customer Wins

OpenAI’s WebRTC Problem (Media Over QUIC)
When you love your tech too much that the realities of using WebRTC doesn’t make sense to you.

📢 Releases

OpenVidu 3.7.0 (OpenVidu)
Oracle cloud support and much needed updates to its SFU versions.

Introducing Audio Connector SDK & Pipecat Serializer for AI Audio Apps (Vonage Developer)
Vonage introduces their AI audio connector and an integration to Pipecat.

Peermetrics at Scale: When WebRTC Monitoring Hits a Million Events a Day (WebRTC.ventures)
Mostly focused on hardening the system in this update.

May Updates v1.5.0: Geolocation, Website Revamp & New Observations (rtcStats)
We’ve added geolocation support across the board.

🤓 From our own posts

• MOQ is lacking a compelling adoption reason (BlogGeek.me)

That’s all for this week. See you next week!

📖 Reading

From the old SkyWay to the new one (X)
PeerJS survived there for 13 years.

⚙️ Technical

WebRTC Live is a monthly webinar series featuring industry guests, hosted by WebRTC.ventures (sponsored)
Join us for the May 13 episode, “WhatsApp Business Calling and SIP” with guest Dan Jenkins of NimbleApe, Everycast Labs, and CommCon.

How OpenAI delivers low-latency voice AI at scale (OpenAI)
Stop everything you’re doing and go read this. More on that in WebRTC Insights, for those who are subscribed.

💼 Use Cases and Customer Wins

Be ready: WebRTC goes grid-wide in Second Life, May 5th, 2026 – Inara Pey: Living in a Modemworld (Modemworld)
Second Life finishes switching from Vivox to its own WebRTC-based voice service.

📢 Releases

Voice Live API now supports WebRTC (Preview) (Microsoft Foundry Blog)
Last week we had WebRTC+WebSocket from Microsoft. This week, WebRTC-only.

🤓 From our own posts

• WebRTC Monitoring: Are You Monitoring Your Servers or Your Service? (BlogGeek.me)

That’s all for this week. See you next week!

📖 Reading

Mozilla patched 271 Firefox bugs found by Anthropic’s Mythos, and says the zero-day era has an expiration date (TNW)
A few related to WebRTC. Expect more such security bugs issued and fixed across all browsers.

Amazon launches an AI-powered audio Q&A experience on product pages (TechCrunch)
WebRTC or WebSocket? That’s the question…

⚙️ Technical

From webrtc dump to root cause in 60 seconds (sponsored)
Drop a webrtc-internals dump into rtcStats. Get an experience score and automated findings in under a minute.

WebTransport Is Now Baseline. Here’s What That Means for Real-Time Media (WebRTC.ventures)
That’s solid analysis. WebTransport support in Safari opens more doors for it.

WebRTC, Compute Pressure and rtcstats (rtcStats)
How to make use and visualize Compute Pressure readings for WebRTC.

Should You Still Consider the AV1 Codec in Your WebRTC Architecture? (WebRTC.ventures)
Some thoughts about the use of AV1, now that a patent litigation is looming about.

From webrtc-internals Dump Files to Production Monitoring: A Practical Migration Guide (HackerNoon)
Why you should collect getStats() data - and how.

💼 Use Cases and Customer Wins

Pepperl+Fuchs Releases Industrial Event Camera and 1D/2D Code Reader (Control Automation)
An industrial event camera that has built in WebRTC support.

LanguageLine Solutions Transforms On Demand Interpretation Services With Vonage Video API (Press release)
Face-to-face video interpretation. LanguageLine makes use of Vonage.

Popup Frees Creators from the Algorithm with Stream’s Livestreaming Infrastructure (Case study)
Live streaming for creators. Popup makes use of Stream.

The Vodia PBX Now Integrates with Jitsi Meet (Press release)
Integrating a PBX with Jitsi Meet to enable video meetings. Makes sense.

Eve Wholesale adds UC Web to platform (Comms Business)
2016 came calling…

📢 Releases

Release Notes for Safari Technology Preview 242 (WebKit)
4 bullets related to WebRTC. That’s almost 4 times the average.

🤓 From our own posts

• WebRTC Monitoring: Are You Monitoring Your Servers or Your Service? (BlogGeek.me)

That’s all for this week. See you next week!

📖 Reading

None this week.

⚙️ Technical

From webrtc dump to root cause in 60 seconds (sponsored)
Drop a webrtc-internals dump into rtcStats. Get an experience score and automated findings in under a minute.

Shipping WebRTC Video From a $10 Microcontroller: Challenges Building the Stream Video ESP32 SDK (Stream)
Down memory lane of my distant past writing code for embedded devices.

Building Real-Time Speech Translation with AI Avatars with Azure Speech Services (Microsoft Foundry Blog)
Why speaker uses WebSocket and listeners use WebRTC is beyond me.

How to Scale Real-Time Video Streaming to 1 Million Viewers in 2026: WebRTC, CDN, and MoQ Architectures (Fora Soft)
Fora Soft is also taking a stab at the scaling a broadcast session issue we’ve seen recently.

Janus RoQs! (Meetecho)
Janus is making more steps towards supporting QUIC (RoQ and MoQ are both there).

💼 Use Cases and Customer Wins

None this week.

📢 Releases

Tagged versions 1.4.1 (multistream) and 0.16.1 (legacy) (Janus discourse)
Vulnerability issues fixed. Go upgrade.

April Updates v1.4.0: Embeddable Viewer, Compute Pressure, and a Brand New Trial (rtcStats)
You can now embed the rtcStats view in your own dashboards.

🤓 From our own posts

• Memory prices, WebRTC and the future of Voice AI (BlogGeek.me)

That’s all for this week. See you next week!

📖 Reading

Escaping the Fork: How Meta Modernized WebRTC Across 50+ Use Cases (Engineering at Meta)
Don’t. Fork. WebRTC. You’re not as big as Meta…

⚙️ Technical

WebRTC Live is a monthly webinar series featuring industry guests, hosted by WebRTC.ventures (sponsored)
Join us for the April 22 episode, “How Experienced Teams Debug and Monitor WebRTC in Production” with guest Justin Williams of WebRTC.ventures

Video AI Avatars: Why We Need to Ditch the Meeting Room Architecture (Gustavo Garcia)
This will greatly reduce complexity and focus on solving video generation limitations we have today.

Why we used STOMP with WebSocket (Behzod Halil)
First time I see STOMP used with WebRTC. I used it ~18 years ago for something similar, before the days of WebRTC.

How We Stream Live Security Cameras to Any Browser Without an AWS Bill That Kills the Business (Nic Mapogha)
Using a local media server instead of a cloud service. Once cameras support WebRTC directly, we will be able to use P2P.

💼 Use Cases and Customer Wins

None this week.

📢 Releases

Fishjam Release Notes: 0.26.0 (Fishjam)
Simulcast, background blur for React Native, and a mobile SDK improvements.

Native VERA 1.1 Release: Meeting Controls and Screen Sharing (Vonage Developer)
Vonage doubles down on their reference application for the Video API platform.

🤓 From our own posts

• Memory prices, WebRTC and the future of Voice AI (BlogGeek.me)

That’s all for this week. See you next week!

📖 Reading

Canadian Astronaut Jeremy Hansen Makes First Space-to-Earth Video Call (World Today News)
At latencies of seconds, this is more akin to streaming than it is to RTC.

⚙️ Technical

webrtc-internals on steroids (sponsored)
10x your troubleshooting and debugging with our powerful WebRTC statistics viewer.

MediaRecorder Video Bitrates (AddPipe)
Always nice to have these posts by AddPipe! They also published one about error handling.

Proton Meet Isn’t What They Told You It Was (Sam Bent)
Using WebRTC isn’t a guarantee of security or privacy. Security and privacy also depends on who is your target audience as well.

💼 Use Cases and Customer Wins

I Couldn’t Call My Mom. So I Built a Video Calling App (Dima Doronin)
Developing “hello world” WebRTC apps is easy these days.

📢 Releases

None this week.

🤓 From our own posts

• Voice AI best practices, based on rtcStats data (BlogGeek.me)

That’s all for this week. See you next week!