Everything below was produced by one Python pipeline from a single JSON script file:
narration, every visual, motion, captions, mix and final render. No editor opened a timeline.
Notes under each sample say exactly what was generated and what was not.
Long-form — The Golden Record
8 min · 1080p · narrated
Rendering. This slot fills when the build completes.
AI-generated
Every visual (Flux-1-schnell) and the narration voice (Deepgram Aura-1). No stock footage, no stock images.
Not AI
The script, written by hand. The music bed is a licensed royalty-free track (Kevin MacLeod, "Anguish", CC BY 4.0, credited in the description).
Human time after the pipeline ran
None. The file the pipeline writes is the file that gets uploaded, with title, description, chapters and tags emitted alongside it.
Shorts
Vertical · burned-in captions · same pipeline, vertical branch
24 s — "This will outlive the human species"
30 s — "Six weeks to speak for all of humanity"
Captions
Three words a group, timed from the narration the pipeline already holds rather than by re-transcribing it, and weighted by character length inside each sentence. Rendered as transparent PNGs and composited, because this FFmpeg build has no libfreetype and therefore no drawtext.