We came to this from the footage side. For years the work was video — analysing it, cutting it, producing things out of it — and most of that work was watching. Someone sits with a timeline and scrubs, because what you need is four seconds somewhere in nine hours, and there is no way to just ask for it.
When models got good enough to actually watch video, we started using them the way you would use a very fast assistant: find the moments that matter, summarise a shoot, draft the report that has to go out on Friday. It worked well enough that it changed how we scoped jobs. It was also a mess — a folder of prompts, a script somebody wrote once and only they could run, three different ways of getting a file in front of a model.
So we systematised it. The prompts became templates. The scripts became a pipeline. The output came back with the same structure every time and with timestamps you could click back to, which is the part that turns an answer into something you can hand to a client. Then we ran our own client work through it, and the slow part of the job stopped being the slow part of the job.
Then the clients started asking about the pipeline itself. Not the report — the thing that made the report. Could they have one for their own footage? That question came up enough times that building it properly stopped being a side project and started being the work.
Video Insight Pro is that, built as a product rather than as a rig we maintain by hand. Upload your footage or connect the drive it already lives on, ask questions of it in plain language, get answers anchored to the second they came from, and turn a library's worth of video into a document somebody will actually read. It is the tool we wanted while we were doing all of this the long way.
We are Launchable AI Inc., based in Ontario, Canada. If you have footage and a question about it, or a pipeline of your own you would like built, tell us about it.