Which video platforms are strong on video and asleep on AI?
The platform you would have picked last year on codecs, scale, and price can be asleep on AI. This edition scores all 17 platforms on two separate axes, video and AI, so you can shortlist in an afternoon instead of burning a quarter on sales calls.
Available now. Launch pricing through July 24.
What this report is
Separate the platform’s video from its AI, before you sign.
You picked a video platform, or you are about to. A year ago that decision came down to codecs, scale, geography, and price. You lined the vendors up on a feature grid and defended the winner in a design review. Then every one of them added AI, and the grid stopped telling the truth.
Open any vendor site today and the same two letters are on the homepage. Some of these platforms rebuilt around serving AI agents: raw audio frames exposed to external models, turn detection, latency budgets measured for a machine instead of a human. Some added noise suppression and a transcription endpoint, put “AI” in the nav, and called it a strategy. On the feature grid they look identical. In production they are not close.
This report is my independent read of the Video API market: who the vendors are, how they score on the dimensions buyers actually weigh, and the part that takes 20 years in this industry to call, where each one is heading. A score tells you who is strong today. The trajectory read is what keeps you from signing a three-year contract with a vendor already walking out the door.
It is more than 180 pages across 16 chapters, plus the tools that make a document that long usable: a searchable membership site, an AI assistant trained on the report, a price calculator, and a side-by-side comparison tool.
Who this is for
Built for the people making the decision, not researching the technology.
Engineering leads & architects
Choosing the infrastructure a real-time product runs on. You want real numbers per platform, named tradeoffs, and a shortlist you can defend in a design review without a vendor deck doing your thinking for you.
CTOs & technical decision-makers
Picking a platform for an AI-augmented product, or running due diligence before a multi-year commitment. Independent analysis you can forward up the chain, and a straight read on who is investing and who is coasting on last year’s roadmap.
Product managers building voice AI
You need to know which platforms are ready for a production agent and which are shipping a good demo.
Vendors covered
17 Video API platforms, scored on the same framework.
Every platform runs through the identical rubric and gets a full profile with a buyer verdict. New in the 3rd edition: SignalWire and Stream.
New in the 3rd edition
Video capability and AI readiness have split.
For a decade, picking a video platform meant weighing codecs, scale, geography, and price. Then the market added AI, and a single blended score started hiding two very different vendors: the one that is excellent at video and asleep on AI, and the one that is the other way around.
So this edition scores every platform on two separate axes: a Video API score out of 100, and a separate AI score out of 100. The two-axis market map puts that gap on a single page, so you can see which platforms rebuilt for AI agents and which bolted on a transcription endpoint and called it a strategy.
AI now carries roughly a quarter of the total score, up from a rounding error in edition 2. Five new dimensions:
Client-side AI
On-device and in-client AI: noise suppression, background handling, and what runs before a frame ever leaves the browser.
Server-side AI infrastructure
Raw audio frames exposed to external models, media routing built for machines, and the latency budget an agent actually needs.
AI framework integration
How cleanly the platform plugs into the open-source agent frameworks: Pipecat, LiveKit Agents, TEN, and Vision Agents.
Voice AI support
What a production voice agent needs: sub-800ms mouth-to-ear latency, turn detection, telephony, and the real cost stack.
AI-friendly development
Can an LLM read the docs and generate working integration code, or does it need tribal knowledge no amount of RAG will fix?
This is not a theoretical exercise. Every platform was tested against it.
Included with every purchase
More than a report. A platform for the decision.
Ask the Report
An AI assistant trained on the full report. Ask a direct question, get the answer with the sources it came from. Like having the analyst on call.
Membership site
Every chapter, tool, and update in one searchable place. Live pages you browse, sections you download, always current.
Price calculator
Drop in your own minutes and workload and see the real cost per vendor: RTC minutes, AI-engine minutes, and standby charges included.
Comparison tool
Put any vendors side by side on any dimension: video score, AI score, or a single criterion.
Every purchase is the report plus these tools. There is no report-only version.
Read the opening chapters before you decide.
The PDF preview includes the complete table of contents and the report’s opening chapters. Use it to see which questions the full report answers, how it separates a platform’s video capabilities from its AI story, and whether the depth is right for the vendor decision in front of you.
The vendor profiles, scores, verdicts, and decision tools are included in the full report.
Pricing
Buy now at launch pricing.
Pay once. Includes the report and every tool. Launch pricing for early readers, expires July 24, 2026. After that it is $2,500 / $3,500. I do not extend launch pricing.
For teams running a vendor evaluation.
- The full report, more than 180 pages
- The membership site with every report tool
- Ask the Report AI assistant (standard access)
- The price calculator
- The comparison tool
For teams that want the analysis and a direct line to me.
- Everything in Standard
- A one-hour call with me to talk through your specific decision
- Unlimited access to the AI assistant
The 3rd edition is available now.
Need an enterprise license or a custom briefing? Get in touch.
Questions
Frequently asked questions.
Is this a video conferencing comparison, like Zoom versus Teams?
No. This is for people building on top of video APIs and SDKs, not choosing an app to hold meetings in. It covers the infrastructure you write code against.
When do I get access?
After purchase, I’ll set up your access to the report and its tools. Launch pricing holds through July 24.
How current is it?
The data reflects mid-2026. Vendors move fast in this market, and the report says so per vendor with the trajectory read. This is the snapshot you build a decision on today, with a flag on who is likely to look different in a year, which is the whole point of reading trajectory and not just the current score.
Do vendors pay to be included or to score higher?
No. Every vendor is scored on the same rubric, and the score is finalized before any commercial conversation happens. A vendor can license a favorable result for its own marketing after the fact, but paying never moves a number. Full detail on the independence section below.
What if a vendor did not respond to you?
Two cases, both stated openly. Vendors too important to omit were scored from public sources and independent research, and their profiles are flagged so you can weight them accordingly. Vendors that did not respond and were not essential were moved to the “what has changed” chapter rather than profiled.
Do I need the tools, or can I just get the PDF?
The tools ship with the report because a document with more than 180 pages needs them. The AI assistant, calculator, and comparison tool are how you get from “all of it” to “the part that answers my question”.
Can I expense this or buy for a team?
Yes. Reply to the order confirmation for an invoice or a team arrangement.
How is this different from the free content on the blog?
The blog is where I think out loud, one vendor or one trend at a time. The report is the scored, structured, side-by-side decision tool: all 17 vendors, both scores, the trajectory read, and the verdicts in one place, on one framework, so you are comparing like with like instead of stitching together 30 blog posts.
Independence
How this report stays independent.
Vendor scores are finalized independently before any commercial conversation begins. Buying a vendor program does not influence current or future scoring. Every vendor runs through the same rubric, self-reported claims are verified against live documentation, and any product I have a commercial interest in is held to the same yardstick and disclosed. That independence is the only reason a score in this report means anything.
What report buyers say
Trusted by the vendors in the report, and the teams evaluating them.
This guide provides a comprehensive overview of the market for real-time video APIs, including technical comparisons that can’t be found elsewhere. The insights provided have helped Agora validate our product roadmap and confirm our commitment to improving the developer experience.
The WebRTC API report is a great resource both for people who are new to WebRTC and for those of us who have been part of the ecosystem for a long time. This comprehensive, unbiased snapshot of vendor features is the only report of its kind.