3D & Generative · Oct 5, 2026
Everyone is posting "Claude made this video in one prompt".
Everyone is posting "Claude made this video in one prompt". by SYGNITO. Watch the 3D & Generative video and visit the original X post.
SYGNITO@SYGNITO
View on XEveryone is posting "Claude made this video in one prompt". I made a whole series of films for SpaceSphere this way. None of them was one prompt. Here is the actual system, step by step. First, what is really happening. The model does not paint a video. It writes a program that knows what the screen should look like at any given moment. A renderer runs that program frame by frame and turns it into an MP4. That is why fixing things is cheap: a title comes in too early, you change one number and render again. 1. Write a brief, not a vibe. One sentence the viewer should remember. Who it is for. How long. And a hard line between facts and creative choices. The model may invent camera moves. It may not invent my product. 2. Zero fake data. My rule for every SpaceSphere film: every number and every screen traces back to a source list. Claude Code captured the real live site in a browser by itself. If a page showed an error or a loading state, the shot was dropped, never faked. Side…
The prompt behind it.
The creator has not shared a public prompt. Visit the original post for more about this work.
Video and prompt by SYGNITO. Original post ↗


