The Continuous Monitoring Problem
A camera generates roughly 30 frames per second. A typical site has 8 to 50 cameras. Even at one frame per second sampled, that is hundreds of thousands of images per day per site. Sending those to a hosted LLM API was never the design intent of ChatGPT, and the cost and latency profile reflect that. Manako runs specialist vision skills locally, on the kind of small machine you can put in a back office, and it runs them at frame rate, 24/7, with no operator watching the dashboard.
This is not a knock on ChatGPT. The tool is genuinely useful for the things it was designed for. It is the wrong shape for site monitoring at any honest budget.
The Hardware and Budget Problem
Most real sites do not have GPU servers and are not going to buy them. The site manager has a fixed budget per location and an existing camera estate they paid for years ago. The question is whether vision intelligence can live inside that budget envelope. With ChatGPT, the per-frame cost is small but the steady-state rate adds up fast. With Manako, the hardware is a small local box and the cost per site is fixed. Predictable budget, no surprise invoice at month-end because someone left a high-traffic camera on.
The Evidence Routing Problem
If you want ChatGPT to tell your forecourt manager that a spill is happening at pump three, you have to build that pipeline yourself. Capture the frame, send it with a prompt, parse the answer, decide if it is actionable, deduplicate, route the alert, attach the clip, log the response. That is a real engineering project and then it has to be kept alive. Manako ships with native integrations for Slack, WhatsApp, Telegram and email. Every capability is also available via REST with webhook callbacks and WebSocket streams. The alert ends up in the channel your team already uses, with the clip attached, with the right person tagged, without an internal project to make it work.
The Data Path Problem
Sending camera frames to OpenAI is incompatible with a meaningful share of real customer environments: fuel retail, healthcare, financial services, defence, anything covered by strict data residency rules, and most EU sites that care about GDPR exposure. Manako keeps everything local. Footage does not leave your device. That is not a feature we added later. It is the deployment model.
Where ChatGPT is the Right Tool
- One-off image analysis. "What is in this photo?" type questions.
- Drafting reports, summarising notes, helping a human reason about a clip.
- Prototyping. Throwing frames at it to see if a use case is even tractable.
- Anything where a human is in the loop, in front of the screen, asking questions.
Conclusion
ChatGPT can look at an image. Manako runs your cameras, 24/7, on the hardware you can afford, and gets the evidence to the person who needs it. Different jobs, different tools. If you are choosing between them for a real deployment, you have not yet defined the deployment.
