My WebRTC predictions for 2025

What’s in store for WebRTC in 2025? Dive into the trends and predictions, from the rise of video use cases to the influence of Generative AI.

Time to look at what we’ve accomplished in 2024 and think what’s ahead of us in 2025 when it comes to WebRTC.

When we look ahead, there are several notable things that glare at us immediately:

1. WebRTC is here, but questions were raised if video use cases are intersting/important enough. Not anymore
2. Interest in the WebTransport+WebCodecs+WebAssembly triplet still exists, but has slowed down a bit. This will take time to mature
3. The recession is still here. There seem to be an end in sight, at least if you believe the news
4. Generative AI is king. And it has come to WebRTC

Last year I was a Senior Director of Product Management at Cyara for all things testRTC. I am still doing that but only part time. The rest? Writing stuff here, courses, workshops, reports and fractional employment consulting on WebRTC and communications (and yes, I can help you if you need me)

If you are interested, you can read my last year’s WebRTC predictions for 2024 😀

Let’s get started here…

The video version

One thing that happened in 2024, was me doing more videos. It started with last year’s WebRTC predictions video, continuing with my monthly chat with Arin Sime, weekly Q&A and programmable video videos.

It meant I was “forced” to do a video for my WebRTC trends and predictions for 2025:

Read on below to get into the details.

Welcome to the era of Gen AI in WebRTC

We are well into the era of differentiation:

Last year, I tried holding on to the era of differentiation in WebRTC… even suggested it was relevant even today. No more.

Everyone is now running after ChatGPT, LLM, OpenAI, … Connecting WebRTC to bots and agentic AI. All the money, attention, focus and resources are there.

The end result? We don’t care about differentiation as much. We care about generative AI and how to bolt it on and into our WebRTC products and services.

Oh, and since I am doing videos now, there’s a video (and an article) on this new Gen AI era of the WebRTC evolution:

What does WebRTC use look like?

We are a third year in a row at 4 times pre-pandemic usage. We can safely say that this is now stable and not going away. Sadly, it isn’t growing either. Will Generative AI in our new era change that? Probably. Time will tell.

Twilio, Programmable Video and the future of video services

Twilio announced sunsetting their Programmable Video service in 2024. Then extended it until 2026. Then retracted it. Twilio will be keeping Programmable Video and focusing on customer engagement. Yay.

This series of decisions has hurt the market:

• It started by stating “video isn’t important enough to run”
• Continued with questioning “can programmable video survive financially”
• Then came the “trusting CPaaS and programmable video vendors is risky”
• And now it is “video is ok, as long as it is customer engagement related, but less if it is group calling”

Somehow, this is confusing would be customers as well as pissing off existing ones – should they stay with Twilio? Build their own? Search for an alternative vendor?

The loser here is the market:

Now that Twilio is back to Programmable Video, we need to see how this will affect everyone else.

How did I do with my 2024 WebRTC predictions?

I spent a considerable amount of time on my predictions in 2024. Let’s see how well I did.

#1 – libWebRTC (and the future of WebRTC)

My money here was on house cleaning. This is roughly what has been happening.

While I tried to argue for work around Insertable Streams, collaboration, AV1 and Voice AI; I believe the majority of the focus has ended up around AV1… with a potential of seeing HEVC in 2025 in some limited shape and form.

Looking at our own WebRTC Insights service, where we track everything and anything WebRTC related. For our recent 4 year review, we looked at the trend of how much we covered in each year:

In terms of issues and bugs, we’re at the lowest point in the past 4 years. This may indicate the stability and robustness of the code – but it may also indicate the level of investment that Google is putting into libWebRTC (=dwindling).

libWebRTC has seen better days than what it had in 2024.

#2 – Machine learning and media processing

Hard to be wrong by stating the obvious…

I indicated that machine learning, LLM and generative AI will be front and center for investments in 2024. And well… It turned out to be true. How (un)surprising.

#3 – The year of Lyra and AV1

Almost.

I decided to take a bet here and risk it a bit.

First “bet” here? While AV1 is still too early a video codec for most, we will still see it in commercial services. And yes, Google Meet is using it. A bit at least.

Second “bet’? That Lyra or a similar AI voice codec will be launched by someone on web browsers. Didn’t happen. We had proprietary AI voice codecs introduced by multiple vendors (Cisco Webex and Microsoft Teams) but none that run in web browsers.

#4 – WebTransport as a real alternative

Nope. Didn’t happen.

I assumed we would see a commercial service using WebTransport instead of WebRTC in production for streaming… I don’t think that happened. And if it did, I am not aware of it.

We’ve also seen a “replacement” of the term WebTransport for media streaming use cases with the term MoQ – Media over QUIC. QUIC is the transport protocol used by WebTransport, so in a way, MoQ is Media over WebTransport… which is what I referred to here.

The end result? We’re now talking about MoQ and had a few people experiment with it, including Lorenzo Miniero from Meetecho in his great series of posts about MoQ (here’s the last one in the series).

One would expect Zoom to show enough interest to run this in production. But what really happened is that Zoom is now on track to support… WebRTC.

But then again, did this happen in 2024? Unfortunately, not yet.

#5 – M&As and shutdowns

Didn’t see this one coming…

I predicted vendors will leave the market or make a small exit by selling or being acqui-hired.

I nailed it – we had Mux shut down its WebRTC video communication service. And Dolby.io keeping its live streaming service but shutting down its video communication service.

But then… we had Twilio double back on their decision to sunset Programmable Video, keeping it alive and switching focus towards customer engagement. Who would have thought… definitely not me 🥴

WebRTC predictions for 2025

Enough about 2024. That’s old news. Lets see what’s going to happen with WebRTC in 2025 😎

#1 – libWebRTC (and the future of WebRTC)

Like last year, I’ll start with libWebRTC, Google’s popular and important library implementing WebRTC.

Sadly, nothing is going to change this year. libWebRTC will stay in maintenance mode. Any improvements made will be because Google needs them for Google Meet.

The number of external contributors to libWebRTC will stay miserably low to non-existent (besides Philipp Hancke I guess).

This state cannot continue forever. One of two things will eventually happen:

1. Some external force other than Google (maybe a few), will step up and start contributing – with Google’s assistance and blessing
2. Developers will move elsewhere (MoQ is the current best candidate)

Both approaches can’t and won’t happen in 2025. It is too early for that due to multiple reasons that I won’t list here just now.

What other scenarios are out there and truly unlikely?

1. Google, putting more investment into libWebRTC directly. There is no incentive for them to do so anymore. They seeded the market. The company is focused on the bottom line now. Why should they? What’s in it for them?
2. Someone will create a new WebRTC library, rivaling libWebRTC in importance and popularity. It might become popular, but it won’t be embedded in Chrome at any point in time, which won’t make it as important as libWebRTC

#2 – Generative AI

Machine learning is nice. Generative AI, Conversational AI and Agentic AI are nicer.

We are still going to have developers use WebAssembly, but that won’t be as interesting. It wasn’t as interesting as I thought it would in 2024, and will likely be less so in 2025.

Why?

Because LLM technologies and generative AI are done predominantly on servers these days – in the cloud. Edge devices don’t have the memory needed to hold the huge models usually – and when they do, they still don’t fit that well into web pages in browsers.

What we will see in 2025 is a continuation of the past couple of months – companies figuring out how best to connect to cloud LLMs and generative AI with voice and video in real time. That means focusing on lower latencies for these specific use cases.

I am focusing here on what companies will do and not the outcomes, because I think we won’t see too much of an outcome in Agentic AI in 2025 yet. It isn’t that these solutions won’t be in production in 2025 – they will. But they won’t amount to a large percentage of the interactions we will have. We are still in early days with this one.

Oh – and it will also mostly be around voice in 2025. A few video vendors may try to “play” this game with generated virtual humans. But the majority will be sticking to voice for now.

#3 – Audio and video codecs are interesting again

In a different way though.

Last year I thought it would be about AV1 and Lyra.

I think differently this year…

We have AV1. Vendors are adopting it. Things will take time here. But it is progressing nicely. As nice as can be expected from such a technology.

But… we see indications of HEVC coming to WebRTC as well 🙀

So I’d like to make my prediction here slightly different – we are going to see quite a few vendors switching from one video codec generation to another in 2025:

• Some will do this from VP8 to VP9
• Others will adopt HEVC
• Many would prepare or even deploy AV1 (Google Meet is running it already in production for many of their sessions)

The migration to AV1 won’t be full. Each vendor using it will also keep other video codecs – for certain scenarios, use cases and devices. AV1 will still be limited in its deployment, but the envelope of the use cases it will encompass will grow each year.

For audio, there are 3 possible scenarios:

1. What was will be again. No progress here, with Opus being our main audio codec. Lyra, Satin and other audio codecs are going to be ignored by the majority
2. AI voice codecs will spurt all around us. Cisco just introduced their own Webex AI codec this year. Others will follow suit with their own proprietary codecs
3. libopus 1.5 or higher gets adopted in libWebRTC, getting AI capabilities to the masses of WebRTC developers just by integration AI improvements right into Opus itself

My wish is for libopus to be integrated. This should be the least effort for the market as a whole to improve things. Philipp says it won’t happen due to the binary size of this libopus version (~3mb). I hope he is wrong…

👉 If Google doesn’t do this for libWebRTC and Chrome, I’d suggest Microsoft, Apple and Mozilla to do this for their visions of libWebRTC in their browsers and stick it to Google, along with publicly announcing such support… Unlike Google, they have nothing to lose

#4 – WebTransport and MoQ will wait a wee bit

A reversal of last year’s prediction…

WebTransport is still quaint. MoQ has a nice sound to it.

None of them are happening just yet.

And they won’t be happening in 2025 either.

These things need time. And I have been a bit too optimistic last year about them.

Likely because what we see now are developers experimenting with the technology, which got me so enthusiastic and hopeful to see something intersting. But it is too early. Hopefully, I’ll change my mind for 2026.

#5 – M&As and shutdowns

2023. Then 2024. And not 2025. It is all one and the same…

The recession is still here. Trump is the next president. Inflation is still high. Interest rates are high. There are indications that the interest rate will continue to be lowered moving forward. But this is far from over. Most countries have too high a debt already. How they are going to solve this is beyond me. It is good that I am an analyst around real time communications and not the economy.

So yes. More vendors are going to be shutting down or getting silently acqui-hired. Or sold at lower rates that they expected just in order to save face. For the most part, investments in and around WebRTC are going to be kept under tight budgets, to be enhanced for those who can tell a compelling Generative AI story for their WebRTC application.

The end result? Still a turmoil. Still a market of buyers.

Welcome to 2025

2025 is going to be exciting, but only for those doing… Generative AI:

• Welcome to the new era of Generative AI in WebRTC
• Sticking around means… integrating with Generative AI
• The rest is going to be mostly housecleaning work in tight budget environments

I can help

There are a couple of things that I am doing that you may want to consider at this point:

1. Upcoming workshop – I’ll be hosting a workshop on Generative AI and WebRTC. Reserve your spot in advance
2. WebRTC Insights – our WebRTC Insights service can keep you up to speed in all the important happenings of WebRTC – technical, market and otherwise
3. Consulting – I moved to part time at Cyara, which means I have actual time to do fractional employment around WebRTC related stuff (think product management, troubleshooting, strategy and marketing). Reach out to me if this is relevant