Your cameras might be "AI ready," but your security stack can still grind to a halt for a simple reason: video has to be decoded, resized, and converted before any AI model can use it. That invisible work is where many surveillance systems hit their limits, even when GPUs are already reserved for inference.
In this episode, we sit down with Isaac Connor, longtime open source developer and one of the driving forces behind ZoneMinder, to explore how CCTV and video management systems have evolved over the past 15+ years.
We discuss the architectures powering everything from small business NVRs to enterprise and government deployments, why so many systems still depend on scaling commodity CPUs, and what changes once AI enters the picture.
We also dive into the codec reality of modern surveillance. H.264 remains manageable, HEVC raises the compute requirements, and AV1 can dramatically increase decoding costs, especially when analytics are added on top.
Because much of the surveillance industry is built around FFmpeg, Isaac explains why adding VPU acceleration can feel like an engine transplant rather than a complete rebuild. We cover high-density video decoding, power and PCIe slot constraints, ONVIF interoperability, and how dedicated video processing can dramatically improve scalability while reducing infrastructure costs.
If you're building or modernizing a VMS, NVR, or AI video analytics platform, this conversation offers practical insight into where today's bottlenecks really are and how to design systems that scale.
Topics covered
– ZoneMinder's origin story and the evolution of open source VMS
– The three tiers of surveillance deployments, from home users to government-scale systems
– Why cloud VMS costs often reflect inefficient video architectures
– The hidden workload between video decode and AI inference
– High-density decoding and why CPUs quickly become the bottleneck
– H.264, HEVC, and AV1 in real-world surveillance deployments
– Power, PCIe slots, and why compact VPUs change server design
– How ONVIF simplifies camera discovery, capability negotiation, and PTZ control
– Modernizing FFmpeg-based video pipelines without rewriting applications
– Where dedicated video processing fits into the future of security analytics
Learn more about NETINT: https://netint.com
Learn more about ZoneMinder: https://zoneminder.com
Interested in joining Voices of Video?
If you're building innovative technology in video, streaming, AI, broadcasting, surveillance, or media infrastructure, we'd love to hear your story. Reach out and let's explore whether your work would make a great episode.
Stay tuned for more in-depth insights on video technology, trends, and practical applications. Subscribe to Voices of Video: Inside the Tech for exclusive, hands-on knowledge from the experts. For more resources, visit Voices of Video.
