I built this because I was bad at it. My first ads had a torch beaming out of the front of a phone screen, a logo stuck on a t-shirt like a decal, and a lifted house that rendered as a garden arch on stilts. The models weren't the problem. I didn't know the tricks yet.
The pro tools are powerful and they hand you the controls, which is exactly what a video person wants. If you're not one, you go and watch tutorials, then pay for a run of generations that come back subtly wrong.
So I did that part. I watched them, read the model docs, and burned the money on bad renders. Then I wrote every fix down and built it into one pipeline. I worked out the whole pipeline, and every trick along the way to make AI videos come out consistent, so you don't have to. You get the version of me that already knows, on the first render.
