The era of the specialist

The era of the specialist is over, and honestly, good.

First Data Scientist, then UI/UX, and lately I've been making AI visuals- both pictures and videos.

Here's what I got wrong first.

Most AI videos are bad because people ask it to make a whole scene. You write a paragraph, hope for the best, and do it again. It's like a slot machine with a monthly fee.

So I stopped writing prompts and started giving it pictures.

I made a still first. A still is easy to judge and easy to throw away. If the light and the crop are right, I have made every decision before anything moves.

Then I animated that still. Six seconds. One job. Nothing else moves.

Then the fun part. I took the last frame of that clip and used it as the first frame of the next one.

That frame had a hairline crack across a mirror. So the second generation didn't need a mirror described to it. It already had one. Cracked. Lit the way I lit it. All it had to do was break the thing.

And lo and behold, within three iterations, I achieved my anticipated result and merged the videos together.

So, personal lesson is to develop such videos upon last frames from the desired output produced- akin to the software development WATERFALL model- damn, once a techie, always a techie :P - what a complex and narrowly understandable analogy.

So: how was it? Be honest. And what say you? Do you agree with my hook?