My Sora Pipeline Is Literally Just Copy/Paste

Sora is out. My pipeline to it is a keyboard shortcut and two browser tabs.

I have a story workshop — a thing I've been building that generates scene-by-scene scripts, shot descriptions, a little narrative scaffolding — and over the last couple days I wired it up to Sora in the most technically ambitious way imaginable: I copy the output from one tab and paste it into the other. That's it. That's the integration. The "pipeline" is my clipboard.

What I did not expect is for it to work this well. The first video out of the generated script was — not bad. Eerie, actually. The kind of eerie where you're not sure if the model understood the scene or just got lucky with the cinematography, and you realize that distinction may not matter.

The story workshop does the thing Sora can't — it holds structure, maintains character continuity, knows what the scene is trying to do before the camera rolls. Sora does the thing the workshop can't — it renders. Together, through the ancient protocol of ctrl-C ctrl-V, they produce something that neither could alone.

There is a serious version of this that involves proper API calls and automated prompt templating and a real pipeline that doesn't depend on me being awake. I'll probably build that eventually. But right now I have a working loop that generates video from a script, and the loop is two windows and a trackpad, and I'm not sure the automation would make the output better.

The first real test of a tool is whether you use it. I used it.