Every few years, a technology comes along that genuinely changes what’s possible in broadcast. IP video did it. NDI did it. Cloud production did it. Each time, the professionals who leaned in early didn’t just adapt — they grew. They built new service offerings, won new clients, and created workflows their competitors couldn’t match.
Visual reasoning is the next one. And the window to be early is right now.What Visual Reasoning Actually Means for Broadcast. Let me be specific, because “AI” has become the most overloaded word in our industry. Visual reasoning is the ability of an AI system to look at a camera feed and understand what it sees — not through pre-programmed rules or trained-on-thousands-of-images object detection, but through natural language. You describe what you’re looking for in plain English, and the system finds it. In real-time. On your existing camera infrastructure.
This is fundamentally different from the computer vision tools broadcast has used for years. Traditional auto-tracking cameras follow faces or motion. They do one thing, and they do it well. But ask them to track “the presenter holding the award” or “the person in the red jersey” or “the product on the left side of the table” and they have no idea what you’re talking about. Vision Language Models — the technology behind visual reasoning — understand both images and language simultaneously. The practical result is that a single system can handle tasks that would have previously required multiple purpose-built solutions, each with their own training data and deployment overhead.
At PTZOptics, we saw this firsthand when we connected Moondream, an open-source Vision Language Model, to our PTZ cameras. The same system that tracks a speaker at a corporate event can track a horse at an equestrian competition, a product on a manufacturing line, or a pastor moving across a stage. No retraining. No new hardware. Just a different text description.Why This Is a Growth Opportunity, Not a Threat
There’s a natural anxiety in our industry whenever automation enters the conversation. But here’s what I’ve observed after two decades of watching broadcast technology evolve: the professionals who thrive aren’t the ones who resist new tools. They’re the ones who use new tools to do things that weren’t possible before. Visual reasoning doesn’t replace camera operators. It opens up productions that couldn’t afford camera operators in the first place. Think about the high school football game that’s never been streamed because there’s no budget for a crew. Think about the three-camera corporate town hall that currently requires a dedicated technical director to switch shots. Think about the house of worship running a single static wide shot because their volunteer team can’t manage PTZ controls during service.
These are all solvable problems now — and somebody is going to solve them. The integrators, engineers, and producers who understand visual reasoning will be the ones building those solutions and getting paid for them.
What’s Possible Today
This isn’t a roadmap pitch. These capabilities exist right now, and broadcast professionals are already putting them to work.
- Intelligent PTZ tracking by description. Connect a PTZ camera to a visual reasoning system and tell it what to follow. “Track the person at the podium.” “Follow the basketball.” “Keep the CEO centered.” The camera moves autonomously based on what the AI sees. We’ve tested this extensively with PTZOptics cameras and the results are production-ready for a growing number of use cases.
- Automated scene switching. Instead of programming complex macros or relying on a human operator, you can define switching logic in natural language. “If someone is at the whiteboard, switch to Camera 2.” “When the room is empty, go to the wide shot.” The system watches the feeds and makes decisions based on what’s actually happening.
- Smart monitoring and alerts. Draw a zone on any camera view and describe what should trigger an alert. “Notify me when someone enters the loading dock after hours.” “Flag when the conference room has been occupied for more than two hours.” This turns every camera into a context-aware sensor.
- Multimodal automation. Combine what a camera sees with what a microphone hears. A person sits down at the desk and says “let’s get started” — that’s a high-confidence signal to begin recording and switch to the close-up. Either signal alone might be a false positive. Together, they’re reliable enough to automate.
- Live scoreboard extraction. Point a camera at a physical scoreboard and the AI reads the scores in real-time, feeding them directly into your graphics system. No dedicated OCR hardware. No manual data entry.
The Integration Story Matters
What makes visual reasoning particularly relevant for broadcast professionals — as opposed to the general AI hype — is that it integrates with the systems we already use. At StreamGeeks, we’ve built working integrations with OBS Studio (via WebSocket), vMix (via HTTP API), and PTZOptics cameras (via their REST API). These aren’t proof-of-concepts gathering dust in a lab. They’re open-source tools that anyone can fork, modify, and deploy. The Visual Reasoning Playground on GitHub includes 17 production-ready tools covering everything from gesture-based OBS control to multimodal audio-video fusion.
The architecture is also designed to be future-proof. We built what we call a “Visual Reasoning Harness” — an abstraction layer that lets you swap out the underlying AI model as the technology improves. The model you use today will be surpassed by something better in six months. Your integration code, your workflow logic, your deployment infrastructure — all of that stays the same.
The Barrier to Entry Has Never Been Lower
Here’s the part that would have been unthinkable even two years ago: you don’t need to be a software engineer to build these systems. AI coding tools like Cursor allow you to describe what you want in plain English and generate working code. The Visual Reasoning Playground provides templates and patterns you can customize. The Moondream API has a generous free tier for experimentation. And the entire ecosystem is open source.
I’ve watched ProAV integrators with no programming background build custom auto-tracking solutions for their clients in an afternoon. I’ve seen worship tech volunteers set up automated camera switching for their Sunday services. The friction that used to exist between “I understand what I need” and “I can actually build it” is disappearing.Where to Start If you’re reading this and wondering where to begin, my advice is simple: go try it.
The Visual Reasoning Playground is live right now at streamgeeks.github.io/visual-reasoning-playground. Every tool runs in your browser — no installation, no downloads. Grab a free API key from Moondream, open any of the 17 tools, and spend twenty minutes experimenting. Use the included sample videos if you don’t have a camera handy.
Start with the Scene Describer — point it at any video and watch the AI describe what it sees. Then try the Detection Boxes tool — type in any object and watch it get highlighted in real-time. By the time you’ve tried three or four tools, you’ll start seeing applications specific to your own work.
The professionals who understand visual reasoning now — who can speak to it intelligently, demo it for clients, and build solutions around it — will have a significant head start as this technology matures. And based on the pace of improvement we’re seeing in Vision Language Models, that maturity is coming faster than most people expect. The tools are free. The code is open source. The only investment is your time and curiosity.
Try the Visual Reasoning Playground →
–About the Author
Paul Richards is the Chief Streaming Officer at StreamGeeks and Co-CEO at PTZOptics. He is the author of Visual Reasoning AI for Broadcast and ProAV, available as a free download at visualreasoning.ai/book. He has authored more than 10 books on broadcast and live streaming technology.
- Visual Reasoning: The AI Capability Broadcast Professionals Can’t Afford to Ignore - February 17, 2026
- How College Sports Are Redefining Broadcast Technology - April 27, 2025
- PTZOptics Hive Launches at NAB 2025, Wins Best Remote Production Software Award - April 10, 2025
