Here is one short video I made, from an empty folder to a finished file. The script, the timings, the pictures, the ones I threw away, and the cost of each step are all left in. You do not have to take my word for any of it.
The anime is the example. The thing worth looking at is how it was made.
I built the machine that runs this, and I kept every log it produced. Search, scoring, selection, generation, assembly, and a check that catches what the other steps miss. A model narrows the options. I decide. On this run I threw out roughly half the pictures I had already picked, because looking at them properly showed they were wrong.
That habit is the whole pitch. I am not showing you that I can make a video. I am showing you that I check my own work, throw away what does not pass, and leave the evidence where you can audit it. Same rule on the four AI systems on my home page: every one names the failure it defends against before the feature it adds.
Short video is cheap to make and expensive to trust. Most suppliers send you the finished file and ask you to believe the rest. This page does the opposite. You get every input, every picture that was rejected, and every timing decision.
You get the same log I keep for myself. That is the difference between a supplier telling you the work is good and being able to check that it is.
Real channel numbers, for calibration: 347 subscribers, 4 videos, best one 1,200 views. Small, and stated rather than dressed up. The pipeline is what scales.