Grading the video
Every run makes a batch of clips. A clip evaluator scores each one, and only the good ones set the bar for the next run.
I did not expect this one to work. I do some video work for partners, and the pipeline spits out a batch of clips on every run. The obvious move is to pick the best one by eye and move on.
Instead I grade them. A clip evaluator scores each clip on brand, lighting, framing, motion, and whether it serves the beat of the story it belongs to. The good ones do not just win this round. They become the bar the next run has to beat.
So the compilations are not just getting made faster. They are getting better, run over run, the same way the proposals did.
It runs real batches through the pipeline today, built and working. I am still tuning it, and I would rather say that than pretend it is finished.
If you want the machine underneath all of this, it is the flagship piece. This is the same machine, pointed at video.